The impact of grid-scale battery storage on electricity market prices

Wholesale power systems have added grid-scale batteries quickly: the largest fleet in this study grew to 15,688.9 MW by 2025, from a base of a few hundred megawatts at the start of the period. Most published estimates of what that has done to prices rest on simulated operation rather than observed dispatch. This thesis works from metered dispatch. Data are assembled for nine markets between 2020 and 2025: ERCOT, CAISO, Great Britain, Germany, Italy and the four Australian National Electricity Market regions. For the seven whose dispatch series pass screening, it measures how far observed wholesale prices differ from a counterfactual built by removing the fleet's metered dispatch. A merit-order curve relating price to residual demand is fitted per market-year on forty equal-count bins and made weakly increasing by a running maximum. Each hour's counterfactual price is read off that curve at the residual demand that would have prevailed without the fleet, and each of five outcomes is the difference between the two series. Thirty-two market-years are retained, giving 160 market-year-outcome cells of a possible 190, with 95 per cent day-block bootstrap intervals. Four of the five outcomes carry one sign in all thirty-two market-years: daily spread, peak shaving and volatility fall, and valley filling rises. Mean price is the exception, falling in twenty-four market-years and rising in eight, and accounting for twelve of the fourteen intervals that cover zero. Magnitudes vary far more than signs. Daily-spread compression runs from 0.04 per cent of a market-year's own mean price in Great Britain in 2020 to 106.57 per cent in CAISO in 2024. Fleet size in megawatts does not order those magnitudes: South Australia held 1,391.9 MW in 2025 against 3,535.9 MW in New South Wales, yet compressed its daily spread by 95.89 per cent of mean price against 20.01. Capacity measured as a share of mean demand ranks the panel far better, at a Spearman correlation of 0.918 against 0.574 for capacity in megawatts, though only three of the thirty-two market-years sit at the rank it predicts. Alongside the price analysis, the fossil share of recorded generation falls in eight of the nine markets — in South Australia from 43.18 to 25.11 per cent — while battery discharge in the tightest one per cent of hours rises in every market with a recorded fleet. In the Australian regions, where frequency-control participation can be observed, batteries supplied up to 98.4 per cent of a region's regulation requirement by 2025. Every price figure is a conditional counterfactual difference rather than an identified causal effect. The curve is fitted to prices that already contain the fleet's operation and no market supplies independent variation in deployment, so the design can neither establish causation nor rule out other hourly influences. The generation and ancillary evidence is descriptive throughout and attributes no change to storage.