{"id":"8809ec5c-400c-469d-898f-043fa2ff45ff","arxiv_id":"2412.03799","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Co-optimizing transmission upgrades and short-duration storage on a 2,000-bus Texas grid shows both are needed to meet 2050 demand, with storage mainly used for peak shaving.","lead":"This paper applies a transmission and storage expansion planning model to a 2,000-bus synthetic Texas grid, using a heuristic to keep the computation feasible. It finds that neither transmission nor storage alone can meet future demand, and that storage's main role is shaving afternoon peak loads.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Storage candidate heuristic may bias the 'neither technology alone' conclusion: Storage Only is never tested with unconstrained siting.","rationale":"The SC heuristic is the weakest point because it directly affects the comparison that supports the paper's headline claim. The paper honestly acknowledges this limitation and provides a concrete counterexample (Fort Worth-Dallas), but the counterexample is only tested for TEP+Storage, not for Storage Only. Since Storage Only is the configuration whose failure is used to prove that storage alone cannot meet future load, the proof is incomplete: the failure could be caused by the heuristic's restricted candidate set rather than by storage's inherent limitations. The paper's own data show that expanding the candidate set materially improves outcomes (25-29% load-shed reduction), so the magnitude of the bias is non-negligible. A focused experiment--expanding the SC set for Storage Only--would settle whether the central claim survives. This does not invalidate the paper's contributions as a case study; the methodology and computational gains are still valuable, and the conclusions are appropriately scoped as conditional on model assumptions. However, the specific assertion that 'neither technology alone can meet future load' is currently supported only under a restrictive heuristic, so the conditional-accept verdict is correct and should be retained.","tokens_in":16634,"tokens_out":3968,"duration_ms":43013,"concrete_test":"Re-solve the Storage Only configuration for 2040 and 2045 using the union (rather than the intersection) of all nodes experiencing curtailment or load shedding across the five representative days as the storage candidate set; if necessary, include all 2000 buses using a decomposition or rolling-horizon method to maintain tractability. If Storage Only achieves zero or substantially lower load shedding in 2040-2045, the conclusion that 'neither technology alone can meet future load' is not robust. As a complementary check, relax the half-charge constraint (Eq. 8) and re-solve TEP+Storage for 2050 to see whether the residual 3343 GWh shedding disappears, which would indicate an additional modeling bias in the co-investment comparison.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that co-investment in transmission and storage is necessary rests on the observation that Storage Only and Transmission Only both shed load by 2040, while TEP+Storage delays shedding to 2050 (Table VI). However, the Storage Only configuration is solved with the SC heuristic (Section III.A), which restricts storage siting to the intersection of nodes experiencing curtailment or load shedding across all representative days. This restriction is not neutral: Section V.G documents that Fort Worth-Dallas, a major load center, was excluded from the candidate set for TEP+Storage in 2050 despite significant shedding, and manually adding it reduced annual shedding by 25-29%. The same bias applies, likely more severely, to Storage Only in 2040 (Houston exclusion). Because the unrestricted 2000-bus storage siting problem timed out (Table II), there is no evidence that a better storage-only plan could not eliminate or greatly delay load shedding. Thus the claim that storage alone cannot meet future load may be an artifact of the heuristic rather than a structural property. The paper's own sensitivity experiments only augment TEP+Storage, not Storage Only, leaving the core comparison untested.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper formulates a mixed-integer linear program for co-optimized transmission and storage expansion planning (TEP+Storage), with line capacity upgrades and short-duration battery siting/sizing, and applies it to the ACTIVSg2000 synthetic Texas grid over 2030-2050. To make the 2000-bus problem tractable, the authors propose a Storage Candidates (SC) heuristic that restricts storage siting to nodes exhibiting renewable curtailment or load shedding in a no-investment recourse run, intersected across representative days. The case study compares TEP+Storage with Transmission Only and Storage Only configurations and reports investment patterns, costs, and load shedding. The paper's central claims are that co-investment in transmission and storage is necessary because neither technology alone can meet future load, that transmission is driven by renewable integration, and that storage is primarily used for peak shaving.","tokens_in":16993,"tokens_out":5114,"duration_ms":47794,"significance":"The 2000-bus spatial resolution substantially exceeds prior TEP+Storage case studies (typically up to 240 buses), and the computational results in Table II show that the SC heuristic converts an intractable instance into solvable ones, which is a practically useful contribution if the heuristic is reliable. The paper is also transparent: Section V.G explicitly documents a failure of the SC heuristic (Fort Worth-Dallas) and reports the resulting load-shedding reduction. However, because the heuristic's candidate restriction is load-bearing for the headline conclusion, the quantitative policy conclusions should be treated as provisional.","major_comments":[{"comment":"The SC heuristic described in Section III.A restricts storage siting to the intersection of nodes with curtailment or load shedding across all representative days. Section V.G shows that this intersection excludes Fort Worth-Dallas in the 2050 TEP+Storage solution despite significant load shedding there, and manually adding those nodes reduces annual load shedding by 25-29%. The spatial conclusion that storage is sited at curtailment/load-shedding nodes is therefore partly an artifact of the candidate selection. More importantly, the Storage Only configuration in Table VI is solved with the same restricted candidate set, and the unrestricted storage siting problem timed out (Table II). The claim that 'neither technology alone can meet future load' is consequently not established for storage alone: a better storage-only plan with an expanded candidate set might delay or eliminate load shedding. Please re-run at least the Storage Only configuration with a union-based or iteratively expanded candidate set, or explicitly limit the claim to storage sited under the SC heuristic.","section":"Section III.A and Section V.G"},{"comment":"The statement in Section V.B that storage installations 'are concentrated in regions with significant solar capacity, as well as in major load centers' and 'align with the paradigm set by the SC heuristic' is circular: the SC heuristic by construction only permits storage at nodes with curtailment or load shedding. As a result, the geographic findings about storage siting do not provide independent evidence about where storage is economically optimal. Please either derive the siting conclusions from the expanded-candidate experiments or acknowledge that the spatial pattern is at least partially predetermined.","section":"Section V.B"},{"comment":"The half-charge residual-energy constraint in Eq. (8) forces storage to end each representative day at 50% state of charge. Section V.F admits that load shedding in the 2050 TEP+Storage case 'could easily be resolved' by violating this constraint. Since the paper emphasizes storage's peak-shaving role and uses representative days, this assumption may systematically undervalue storage and bias the co-investment conclusion. A sensitivity analysis that relaxes Eq. (8) or uses longer storage operational horizons (as in refs [35], [36]) should be reported.","section":"Section V.F and Eq. (8)"},{"comment":"The Storage Only configuration has no reported results for 2045 and 2050 in Table VI, and several rows of Table VII are incomplete due to timeouts, as explained in Section V.G. The conclusion that 'neither technology alone can meet future load' rests on the Storage Only case shedding load in 2040; without later-year results, the comparison with TEP+Storage and Transmission Only is asymmetric. Please report the timed-out instances explicitly (e.g., with optimality gaps and best bounds) or justify why 2040 results suffice for the claim.","section":"Table VI and Section V.G"}],"minor_comments":[{"comment":"In Constraints (13)-(16), the index r is used for the storage operation variables, whereas the model formulation and nomenclature define the set S with index s; this should be made consistent.","section":"Constraints (13)-(16)"},{"comment":"Figure 6 is missing subfigure (l) for the 2045 Storage Only case and (o) for 2050 Storage Only, despite panel labels in the caption; the visual comparison for Storage Only is therefore incomplete.","section":"Figure 6"},{"comment":"Table VI does not include 2045 and 2050 rows for Storage Only; if these instances timed out, indicate this explicitly with an entry such as 'timed out' rather than omitting the rows.","section":"Table VI"},{"comment":"The column header 'SC # max investments' is unclear; it should be split into e.g., 'Number of storage candidates' and 'Number of lines with maximum upgrade level'.","section":"Table VI header"},{"comment":"In Table VII, the 150% storage-cost case is missing the 2045 and 2050 rows without a note until Section V.G; add a table note so the incomplete rows are interpretable.","section":"Table VII"},{"comment":"The paper states in the abstract that co-optimization 'enhances grid reliability,' but the TEP+Storage configuration still sheds 3343 GWh in 2050 (Table VI); consider qualifying the reliability claim as a reduction in load shedding relative to single-technology investments rather than elimination.","section":"Abstract and Section V.A"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is honest about the SC heuristic's limitations, but the paper's headline contribution ('neither technology alone can meet future load') is not supported by the experiments as designed. The core computational idea is publishable, but the claims need substantial reworking or additional experiments. For a journal, I would recommend major revision rather than reject, because the issue is fixable by expanding the candidate set and re-running the Storage Only cases."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth reading: this is the first TEP+storage co-optimization at 2000-bus resolution that I know of, and it comes with an unusually candid discussion of its own limitations. The storage-candidates (SC) heuristic is simple but makes the problem tractable, and the authors show they know it is imperfect—Section V.G is refreshingly direct about the Fort Worth-Dallas miss and the 25–29% load-shed reduction from adding those nodes. That honesty is the paper's best feature.\n\nWhat is actually new: the scale, not the formulation. The model is a direct adaptation of the linearization in [21], and the SC heuristic is more of an engineering rule than a theoretically grounded method. Still, getting a 2000-bus case to solve within 72 hours and reporting the compute times is a legitimate contribution for the planning community. The geographic results—transmission driven by renewable expansion in the west, storage for peak shaving in load centers—are plausible and well-illustrated.\n\nThe stress-test concern holds up. The headline claim that \"neither technology alone can meet future load\" is built on a comparison where Storage Only is solved with the SC heuristic, which restricts storage siting to nodes with curtailment/shedding on all representative days. The authors never test Storage Only with a broader candidate set, so we don't know if a better storage-only plan would eliminate or delay shedding. The paper's own experiments for TEP+Storage show that excluded load centers matter; the same logic applies to Storage Only, likely more severely. The claim is not airtight. I'd soften it to \"under the SC restrictions, storage alone did not meet load.\" The co-investment benefit is still supported, but the \"neither alone\" assertion goes beyond the evidence.\n\nOther soft spots are minor in comparison: no code or data released, uniform load/renewable scaling, only five representative days, and the half-charge end-of-day constraint (which the authors themselves note may understate storage value). These are standard simplifications, but they mean the geographic investment patterns should be read as illustrative rather than prescriptive.\n\nThis paper deserves a serious referee. It is not a methodological breakthrough, but it is a competent, honest case study that pushes the state of practice in scale. I'd ask the authors to either run Storage Only with a larger candidate set or rewrite the conclusion to match what the experiments actually show. For a reader working on TEP or storage expansion, it's a useful reference and a good starting point for future work on dynamic candidate selection.","headline":"Solid 2000-bus TEP+storage case study, but the storage-candidate heuristic undercuts the 'neither technology alone' claim.","tokens_in":17369,"tokens_out":1593,"would_cite":true,"duration_ms":18982,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A 2,000-bus synthetic Texas case study finds that co-investing in transmission upgrades and battery storage is necessary to meet future demand, while either technology alone begins load shedding in 2040.","keywords":["transmission expansion planning","energy storage siting","storage candidates heuristic","renewable integration","high spatial resolution","co-optimization","peak shaving","synthetic Texas grid"],"falsifier":"Re-run the TEP+Storage model with the SC candidate set expanded to the union of load-shedding and curtailment nodes across days, or with Fort Worth-Dallas nodes added manually as the paper does; if the expanded set eliminates or markedly reduces the 2050 load shed at equal or lower total cost, the intersection rule is a true bottleneck. A smaller-scale test: solve the full TEP+Storage problem on a 24- or 53-bus system and compare against SC-restricted solutions to check whether the heuristic can provably miss optimal sites.","tokens_in":16404,"feed_emoji":"⚡","tokens_out":6096,"duration_ms":55703,"temperature":0.7,"pith_summary":"The paper argues that on a 2,000-bus synthetic Texas grid, planning transmission upgrades and short-duration battery storage together is necessary to meet growing demand under high renewable penetration: investing in either technology alone starts shedding load in 2040, while the co-optimized plan delays load shedding to 2050. To make this scale tractable, the authors propose a storage-candidates heuristic that restricts battery siting to nodes already showing curtailment or load shedding in a no-investment simulation. The case study finds that transmission upgrades are the main tool for moving renewable energy out of congested western regions, while storage primarily shaves afternoon peaks in load centers and appears only in later investment periods. The paper also documents a limitation of its own heuristic: it can omit good storage sites, and Fort Worth-Dallas is excluded despite significant load shedding there in 2050.","feed_headline":"Neither transmission nor storage alone can meet future load","feed_subtitle":"A 2,000-bus Texas study shows co-optimizing line upgrades and batteries keeps the grid reliable a decade longer.","key_machinery":"The storage candidates (SC) method is the mechanism that carries the argument: solve the operational recourse problem with no transmission or storage investments for each representative day, identify the nodes that experience load shedding or renewable curtailment, and take the intersection of those node sets across all representative days as the allowed storage siting set. This reduces the storage siting decision from 2,000 binary variables to roughly 50-150 candidate nodes, cutting the total number of binaries from 50,000 to a few thousand and bringing solve times from over 72 hours down to under 5 hours. The intersection rule is also the source of the main suboptimality the paper identifies, since a node that sheds load on only some days is excluded.","core_discovery":"The central claim is that transmission and storage are complementary investments in a high-renewable grid: line upgrades let zero-cost renewable generation reach load centers, while batteries store energy from low-load hours to meet the afternoon peak. On the synthetic Texas network, the co-optimized TEP+Storage plan installs no storage until 2040, then builds 109.6 GWh by 2050, while the Transmission Only and Storage Only configurations each begin load shedding in 2040. Co-investment delays but does not eliminate load shedding: the model still sheds 3343 GWh in 2050, which the authors attribute largely to the storage-candidates method missing the Fort Worth-Dallas area. The authors state this as evidence that neither technology alone can meet future load and that co-optimization is necessary to keep costs low.","pith_inferences":["An iterative version of the SC method that adds newly revealed load-shedding centers to the candidate set after each investment period would likely recover part of the missed savings; the paper's own manual experiment adding Fort Worth-Dallas storage cut 2050 load shed by 25-29%.","The end-of-day half-charge constraint on storage may be inflating the reported 2050 load shedding, because the paper notes the shed could be reduced by discharging batteries below that residual level.","A sub-hourly operational model would likely recommend more storage than this hourly model, since the paper expects the hourly timescale to miss storage's value for ramping and ancillary services."],"forward_implications":["Co-investment delays the onset of load shedding from 2040 to 2050 on the studied system, although it does not eliminate it.","Transmission upgrades concentrate in high-renewable western regions and dominate early investments, while storage appears later and is placed in solar-rich and load-center nodes.","Storage investment remains necessary across cost variations from 75% to 150% of the baseline, indicating the modeled system needs storage regardless of price.","The storage-candidates heuristic makes a 2,000-bus TEP+Storage model solvable in hours, where the unrestricted model times out after 72 hours."],"supporting_citations":[{"why":"Supplies the linearization that models storage efficiency while preventing simultaneous charging and discharging, which the formulation adapts.","marker":"[21]"},{"why":"Provides the evidence that low spatial resolution causes the largest investment errors, motivating the high-resolution case study.","marker":"[9]"},{"why":"Describes the methodology used to create the synthetic 2,000-bus Texas network used in all experiments.","marker":"[28]"},{"why":"Supplies the generation and load projections used to scale the system from 2022 to 2050.","marker":"[29]"},{"why":"Provides transmission line capacity upgrade costs used in the investment objective.","marker":"[26]"},{"why":"Provides the utility-scale battery storage cost estimates used to set storage power and energy rating costs.","marker":"[27]"},{"why":"Offers a security-constrained co-planning baseline with higher temporal resolution that this paper extends to higher spatial resolution.","marker":"[12]"},{"why":"Co-plans transmission and merchant storage, giving context for why this paper models storage differently and omits merchant ownership.","marker":"[19]"}],"fun_headline_variants":["Co-optimizing lines and batteries delays load shedding","Texas grid study: co-invest in transmission and storage","Neither transmission nor storage alone meets future load","Batteries and lines: complementary, not interchangeable"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the nodes which show curtailment or load shedding in every representative day of a no-investment simulation already contain all the storage locations a good plan could need; the paper's own Fort Worth-Dallas counterexample shows this assumption can fail.","fun_headline_variants_meta":{"raw":{"variants":["Co-optimizing lines and batteries delays load shedding","Texas grid study: co-invest in transmission and storage","Neither transmission nor storage alone meets future load","Batteries and lines: complementary, not interchangeable"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000302,"raw_usage":{"total_tokens":1729,"prompt_tokens":926,"completion_tokens":803,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":542,"completion_tokens_details":{"reasoning_tokens":741}},"tokens_in":542,"tokens_out":803,"duration_ms":8440,"temperature":1.0,"reasoning_tokens":741,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T22:04:31.125101+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the TEP+Storage model with the SC candidate set expanded to the union of load-shedding and curtailment nodes across days, or with Fort Worth-Dallas nodes added manually as the paper does; if the expanded set eliminates or markedly reduces the 2050 load shed at equal or lower total cost, the intersection rule is a true bottleneck. A smaller-scale test: solve the full TEP+Storage problem on a 24- or 53-bus system and compare against SC-restricted solutions to check whether the heuristic can provably miss optimal sites.","supporting_citations":[{"cited_title":"Optimal energy storage siting and sizing: A wecc case study,","cited_arxiv_id":null,"evidence_quote":"Supplies the linearization that models storage efficiency while preventing simultaneous charging and discharging, which the formulation adapts."},{"cited_title":"Quantifying the Impact of Energy System Model Resolution on Siting, Cost, Reliability, and Emissions","cited_arxiv_id":"2406.16924","evidence_quote":"Provides the evidence that low spatial resolution causes the largest investment errors, motivating the high-resolution case study."},{"cited_title":"Grid structural characteristics as validation criteria for synthetic networks,","cited_arxiv_id":null,"evidence_quote":"Describes the methodology used to create the synthetic 2,000-bus Texas network used in all experiments."},{"cited_title":"Transmission expansion planning including tcscs and sfcls: A minlp approach,","cited_arxiv_id":null,"evidence_quote":"Provides transmission line capacity upgrade costs used in the investment objective."},{"cited_title":"Cost projections for utility-scale battery storage: 2023 update,","cited_arxiv_id":null,"evidence_quote":"Provides the utility-scale battery storage cost estimates used to set storage power and energy rating costs."},{"cited_title":"Co-planning of investments in transmission and merchant energy storage,","cited_arxiv_id":null,"evidence_quote":"Co-plans transmission and merchant storage, giving context for why this paper models storage differently and omits merchant ownership."}],"review_version":1}