Why efficient heat pumps do not always save money: field performance and delivery economics

1. A recovering market does not make every project economic

IEA’s July 2026 Heat Pump Monitor describes double-digit European growth in 2025 alongside stabilising global building heat-pump sales. Upfront spending and electricity prices relative to fossil fuels remain deployment constraints.[1] Market direction and a household investment decision therefore need separate tests.

Ask how much a system saves each year at equivalent comfort, hot-water demand and weather, and how much extra the household must pay. A nameplate comparison can omit emitters, storage, electrical work and commissioning; a bill-only comparison can mistake a comfort change for an efficiency gain.

2. What field measurements establish—and what they do not

Fraunhofer ISE’s WP-QS project monitored 77 residential installations. Its main 2024 evaluation covered 49 air-source systems with SPF 2.6–4.9, averaging 3.4, and 13 ground-source systems with SPF 3.6–5.4. Building age did not correlate with efficiency in this sample.[2] A 2026 proceedings abstract gives a ground-source mean of 4.3 and defines SPF as generator efficiency for combined space heating and domestic hot water.[3]

This supports the feasibility of good performance in older buildings, while leaving retrofit needs to be assessed individually. Supply temperature and building conditions matter when transferring the finding. The plotted ranges are observed minima and maxima, not confidence intervals; heat sources were not randomised, so the mean difference is not a causal gain from changing source.

3. Monitoring creates value when deviations become actionable

A 2025 Nature Communications study analysed 1,023 systems from one manufacturer. Efficiency classification used 708 systems with adequate fits: 612 air-source and 96 ground-source, of which 105 and two respectively fell below the applied thresholds. It covers space heating and uses device-derived energy estimates rather than wholly independent metering.[4]

Its standardised metric should not be pooled with annual SPF above. A more useful product implication is a closed service loop: detect deviations, explain plausible causes, arrange installer review, then verify comfort and consumption. A warning alone is not an energy-saving service; false alarms and unresolved alerts can add customer work.

4. Delivery evidence: the product includes emitters and aftercare

The UK Electrification of Heat installation report records 742 systems: 93% of participating installed homes needed at least one radiator replaced and 81% needed new thermal storage. Published in 2022, these are historical demonstration-project requirements, not current market rates or prices; contractor differences also limit some comparisons.[5]

Delivery contractor Warmworks reports 305 installations in southeast Scotland, followed by monitoring and optimisation support.[6] This is an example of delivered installations and service. Its impact descriptions remain company disclosures rather than independently audited savings guarantees. Field-performance variation explains why service capability belongs in procurement.

Break quotations into an accountable service: heat-load and emitter assessment, equipment and hydraulics, installation and reinstatement, commissioning, metering and first-season review. Assign responsibility before comparing complete-system offers. Cheap equipment in a mismatched system can shift the initial saving into continuing energy and complaint costs.

5. The fuel-cost boundary: the price ratio must be below SPF/η

For equal annual useful heat Q, heat-pump electricity costs Q×pe/SPF and boiler fuel costs Q×pg/η. Electricity and gas-input prices pe and pg use the same currency per kWh; η is seasonal boiler efficiency. Lower heat-pump fuel cost requires pe/pg < SPF/η. Gas calorific-value conventions must match η, and heat-pump electricity must include auxiliaries appropriate to the comparison.

With Q=12,000 kWh of useful heat/year, pe=0.30, pg=0.08, SPF=3.4 and η=0.90, annual fuel costs are 1,058.82 and 1,066.67 currency units, a difference of only 7.84. The threshold price ratio is 3.78 versus an input ratio of 3.75. Seasonal efficiency above three can still yield almost no fuel-bill advantage in this hypothetical tariff structure.

Changing only pe to 0.20 produces annual savings of 360.78; keeping the original tariff but raising SPF to four yields 166.67. These are scenarios, not promises. Time-of-use operation can change the weighted electricity price, but also temperatures, losses and SPF; the cheapest tariff and best efficiency cannot be combined without checking that the control strategy can deliver both.

6. Fuel savings still need an investment case

The model separates net incremental installation cost ΔK from annual fuel savings S. Simple payback ΔK/S is meaningful only for S>0. At the default ΔK=4,000, payback is about 510 years, far beyond a normal equipment decision horizon; with pe=0.20 it is about 11.1 years. This is no lifetime prediction: a thin operating advantage cannot support a large upfront premium.

Compare complete alternatives at the same replacement date and subtract only available grants from ΔK. The tool excludes discounting, maintenance, replacement, gas standing charges, finance, cooling value and carbon costs. Avoided standing charges count only if the gas connection is actually removed; continued gas cooking can prevent that saving. Zero or negative fuel savings display no positive payback rather than a negative number of years.

7. Carbon and cost require explicit, separate boundaries

For grid-emission factor ge and gas-combustion factor gg, lower operational emissions require ge/SPF < gg/η for equal useful heat. Cost and carbon need not move together: tariffs include taxes and market structure, while emissions reflect energy supply. Annual averages cannot determine the marginal emissions of shifting a particular operating hour.

This is still not a lifecycle comparison. Equipment production, refrigerant leakage and recovery, drilling or installation materials, replacements and future grid changes need separate inventories. German field findings do not establish a Chinese household’s emissions reduction, and this fuel-cost tool is not a lifecycle-cost model. Missing local evidence remains an open requirement.

8. Adoption needs a verifiable delivery commitment

Priority candidates can meet comfort needs at low supply temperatures, already have suitable emitters, face a sensible replacement window and retain operating-cost headroom under local tariffs. Tight space, noise, electrical-capacity or disruption constraints require feasibility work first. Building age can guide questions but cannot replace a site assessment.

Ask four acceptance questions: are design heat load and supply temperature documented; does metering cover promised electricity boundaries; will first-season performance be reviewed alongside weather and comfort; and who investigates and recommissions deviations? Product and design-management teams can put these into a service process more usefully than an unbounded maximum-savings percentage.

Heat-pump value resides both in the machine and in the capability to connect a building, its energy prices and continuing service. When performance commitments become measured results under equivalent conditions, customers can judge whether the heating solution fits their needs.

1. 市场复苏不等于每个项目都划算

IEA 于 2026 年 7 月发布的 Heat Pump Monitor 指出，2025 年欧洲市场实现两位数增长，但全球建筑热泵销售整体趋于稳定。它同时把初始费用和电力相对化石燃料的价格，列为部署的重要约束。[1] 因此，市场方向与单个家庭的投资判断需要分开。

一个实用问题是：在同样的室温、热水需求和气候条件下，新系统每年少花多少钱，家庭为此需要多支付多少？如果只比较设备铭牌，散热器、水箱、配电和调试可能在报价之外；如果只比较账单，室温变化又可能被误认作效率收益。

2. 实测告诉我们什么，又没有告诉我们什么

Fraunhofer ISE 的 WP-QS 项目监测了 77 套既有住宅热泵。2024 年主要评估中，49 套空气源系统的 SPF 为 2.6–4.9、平均 3.4；13 套地源系统为 3.6–5.4。项目报告没有发现建筑年代与效率的相关性。[2] 2026 年会议论文摘要补充地源平均值为 4.3，并把 SPF 明确为供暖与生活热水合计的热源设备效率。[3]

这支持“旧建筑也可能运行良好”，不支持“所有旧建筑都无需改造”。样本的供水温度与建筑条件才是转移结论时应检查的变量。图中区间是已观测最小值到最大值，不是置信区间；两种热源没有随机分配，均值差不能当作更换热源的因果收益。

3. 数字监测的价值，是找到可处理的偏差

2025 年 Nature Communications 研究分析了单一制造商的 1,023 套系统；效率分级实际使用 708 套拟合合格系统，包括 612 套空气源和 96 套地源。模型标准化 SCOPreal 低于所用门槛的分别有 105 套和 2 套。它聚焦空间供暖，输入能量由设备传感器推算，并非全部独立热量表测量。[4]

因此，这篇研究的标准化指标不与上一节的年度 SPF 合并排名。对产品团队更有用的启发，是把监测设计成闭环：识别偏差、解释可能原因、安排安装商复核，再确认舒适度与能耗是否改善。一个红色告警本身不是节能服务，误报和无人处理的告警还会增加用户负担。

4. 交付案例：真正的产品包含散热器与售后

英国 Electrification of Heat 的安装报告记录了 742 套系统：已安装家庭中，93% 至少更换一个散热器，81% 需要新的蓄热装置。报告发表于 2022 年，说明的是当时示范项目的改造工作量，不是今日全市场比例或报价；不同承包商的方法也限制部分结果的直接比较。[5]

交付商 Warmworks 披露，它在苏格兰东南部完成 305 户安装，之后跟踪运行并提供优化支持。[6] 这构成实际安装与服务的案例；其效果描述仍是企业自述，不能视为独立审计的节费保证。研究发现的运行差异，恰好解释为什么服务能力需要进入采购范围。

报价应按一套可验收的服务拆解：热负荷与末端核查、设备与水力系统、施工与恢复、调试、计量和首个供暖季复核。各项明确责任方，才能比较两份“整套”报价。把便宜设备装进不匹配的系统，可能把节省的采购费转化成长期电费和投诉成本。

5. 省钱边界：电气价格比必须低于 SPF/η

对同样的年度有用热量 Q，热泵电费为 Q×pe/SPF，燃气锅炉燃料费为 Q×pg/η。pe 与 pg 分别是每 kWh 电力与燃气输入的同币种价格，η 为锅炉季节效率。热泵燃料费较低的条件是 pe/pg < SPF/η。燃气热值口径必须与 η 一致；热泵系统电耗边界也应包含比较所需的辅机。

示例设 Q=12,000 kWh热/年、pe=0.30、pg=0.08、SPF=3.4、η=0.90。热泵与锅炉年燃料费分别约 1,058.82 和 1,066.67，差额仅 7.84 货币单位；边界价格比为 3.78，而输入价格比已达 3.75。高于 3 的季节效率，在这个假设价格结构下仍几乎没有燃料费优势。

若只把电价改为 0.20，年节省变为 360.78；若维持原电价、把 SPF 提至 4，年节省为 166.67。这些是同一公式的情景，不是承诺。分时电价可改变加权购电成本，但移峰也可能改变水温、损耗和 SPF；不能同时取最便宜电价与最理想效率，却不验证控制策略能否兼得。

6. 从燃料费优势到值得投入，还差一张现金流表

交互模型将净增量安装投入 ΔK 与年燃料节省 S 分开。只有 S>0 时，简单回收期 ΔK/S 才有意义；默认 ΔK=4,000 时回收期约 510 年，远超通常设备决策期限。将电价改为 0.20 后约为 11.1 年。这不是寿命预测，而是提醒读者：很薄的运行优势不足以支撑显著的前期溢价。

ΔK 应比较相同更换时点的完整方案，并扣除实际可获得的补助。模型未计折现、维修、替换、燃气固定费、融资、制冷价值或碳成本。若燃气连接确实取消，固定费节省可另行加入；如果仍用燃气做饭，就不能默认这笔节省。零或负的燃料节省会显示“无正向回收”，而不是负年数。

7. 减碳与省钱，应使用两套清楚的边界

同样的运行热量，若电网排放因子为 ge、燃气燃烧因子为 gg，运行减排条件是 ge/SPF < gg/η。这个式子解释了省钱和减碳为何不必同时发生：价格包含税费与市场结构，排放取决于能源组合。年度平均因子也不能回答某个时段移峰的边际排放问题。

这仍不是生命周期比较。设备生产、制冷剂泄漏与回收、钻井或安装材料、替换次数和电网未来变化，都需要另设清单。本文不把德国项目的结果套用为中国家庭的减排百分比，也不把节费模型输出称作生命周期成本。当地数据缺失时，应该保留待验证项。

8. 值得采用的条件，是一份能验证的交付承诺

优先评估的场景，是在舒适度不下降的前提下能以较低供水温度供热、已有合适末端、替换窗口合适，且价格比给运行成本留下余量的建筑。空间、噪声、供电容量或施工扰动约束很强时，应先评估可行改造方案。建筑年代可以提示调查方向，不能替代现场核查。

交付验收可围绕四个问题：是否记录设计热负荷与供水温度；计量是否覆盖承诺的耗电边界；首个供暖季是否在记录天气与舒适度的同时复核性能；偏差出现后谁负责复查与调试？对于产品与设计管理团队，这些比一个没有边界的“最高节省”百分比更适合写进服务流程。

热泵的商业价值不只存在于压缩机里，也存在于把建筑、能源价格与售后服务接起来的能力里。当效率承诺能够转化成同条件的实测结果，客户才能判断自己购买的是合适的供热方案。