To gain a clear understanding of the current state and to identify the driving factors influencing the power generation performance of MFCs in China, a comprehensive database was built by searching the CNKI database using the following terms: “(microbial fuel cell) AND (power generation performance OR coulombic efficiency OR current OR voltage OR power density OR current density)” in the keyword, abstract, and title fields to retrieve relevant literature (data were updated prior to 30th November, 2024). However, the inclusion of only Chinese publications may not fully represent all MFC studies conducted in China. Nevertheless, our dataset remains meaningful, as many researchers publish papers in both Chinese and English, and some papers may share the same data derived from a common series of experiments, especially regarding device configurations, reaction conditions, and substrate characteristics. To ensure the quality of our dataset, we excluded publications that had not been peer-reviewed (e.g., academic dissertations). Initially, we selected 883 publications, which were carefully screened based on specific criteria to ensure their suitability and relevance. First, the selected MFC studies had to be experimental, excluding review papers. Secondly, they had to report at least one index of power generation performance (e.g., voltage, power density, or coulombic efficiency). Finally, these studies had to provide at least one index related to MFC device configurations, reaction conditions, or substrate characteristics. After rigorous screening and evaluation (Figure 1), our final dataset comprised 10,826 cases drawn from 186 publications (Appendix I). For each publication, we collected data on power generation performance (current, voltage, power density, current density, and coulombic efficiency) and three categories of potential influential factors: device configurations (e.g., anode surface area, chamber volume, and cathode surface area); reaction conditions (e.g., ambient temperature, pH, and external resistance); and substrate characteristics (e.g., pre-treatment type and operating phase). Data on MFC power generation performance and the potential driving factors were either directly collected from tables and manuscripts or indirectly extracted from figures using an online tool (https://plotdigitizer.com/app). We grouped the operating phases of substrates into three categories: liquid, solid, and fluid. We also classified the pre-treatment types into biological, chemical, physical, and multiple methods (e.g., biochemical and physical chemical). In addition, to enable comparisons with international counterparts, we also collected data from relevant meta-analyses or review papers (Appendix II).

First, to compare the results between our study and other meta-analyses or review papers, t-tests (for voltage and reaction duration) and one-way ANOVA (for power density, temperature, and coulombic efficiency) were conducted based on the number of studies with available data. For one-way ANOVA, post hoc multiple comparisons were performed using Tukey's test (assuming equal variance) and Dunnett's T3 test (assuming unequal variance) when differences were significant (P < 0.05). Then, the sample size, mean, median, extreme value, coefficient of variation (CV), and other indices were calculated for all collected data to examine general patterns. All potential influential factors were grouped into three categories: device configurations, reaction conditions, and substrate characteristics. In order to test the effects and strengths of each potential influential factor on each power generation performance index, t-tests (for reaction chamber type and substrate pre-treatment) and one-way ANOVA (for substrate type) were conducted for categorical variables, and correlation analyses were conducted for continuous variables. Pearson and Spearman correlations were used for parametric and non-parametric data (the latter only for current), respectively. We also tested non-linear relationships and compared them with linear relationships. Non-linear models were selected when R² values were substantially higher than those of linear models. For each group of potential, influential categorical factors, if a focal power generation performance index was significantly correlated with two or more continuous factors, a multiple linear regression model was conducted to identify the main contributors. Finally, a generalized linear model was used to investigate whether there were interactions between categorical variables and continuous variables identified by the multiple linear models. Before conducting these analyses, the normality, outliers, and homogeneity of variance of all data were examined, and data were transformed to meet the assumptions of normality and homogeneity (see Supplementary Tables S1, S2 for details of data transformation). Indices collected from fewer than three studies or with fewer than 30 cases were excluded from all tests. All statistical analyses were conducted using IBM SPSS Statistics 26.0, and all figures were generated using Origin 2025b.

We found generally lower power density, voltage, and reaction duration but higher operating temperature and coulombic efficiency in our study compared to other studies (Figure 2). More specifically, the power density in our study (59.95 ± 6.11 mW/m²) was significantly lower than that in other studies (Figure 2c). The volumetric power density in our study (2.56 ± 2.03 W/m³) was only 60.2 % of that reported by Dowdy et al. (2018). However, it showed no significant difference compared to the values reported by Bird et al. (2022) and Agrahari et al. (2022). Coulombic efficiency significantly differed among studies (Figure 2f); our study reported a value of 9.30 ± 0.76%, which was 18.4 and 33.1% higher than those reported by Dowdy et al. (2018) and Bird et al. (2022), respectively, but showed no significant difference from Zafar et al. (2022). Moreover, significant differences in voltage were observed among studies (Figure 2e), with our study recording 182.69 ± 2.16 mV, which was 19.8% lower than the values reported by Amin et al. (2022). Regarding reaction conditions, the reaction duration in our study (78.04 ± 4.31 h) was lower than that reported in the study by Bird et al. (2022) but showed no significant difference compared to Uddin et al. (2021). Our study also reported relatively higher temperatures than those in the other studies (Figure 2a).

The general information on the mean, median, maximum, minimum, and coefficient of variation (CV) for all data is shown in Table 1. The majority of CV values fell within the range of 20–40%, with the exception of power, which had a CV as high as 306.61%, and ambient temperature, which had a CV of only 4.32%. Power density and current density also exhibited relatively high CV values (87–89%), while external resistance showed a relatively low CV (11.55%). CV values were generally high for power generation performance indices (>40%), while they were typically low for device configurations (<40%). The majority of data showed relatively small variances between mean values and median values, except for internal resistance, for which the mean was 15.1% higher than the median value. However, the relatively low CV values may be due to the data transformations applied.

The characteristics of MFC devices were influenced by reaction chamber type, substrate pre-treatment, and substrate operating phase (Figure 3). Dual-chamber MFCs exhibited 4.7–36.8% higher values for the majority of device characteristics compared to single-chamber MFCs. However, dual-chamber MFCs had 3.0 and 12.1% lower electrode surface area and reaction duration, respectively, than single-chamber MFCs. Reaction duration, pH, and internal resistance were 12.9%, 3.7%, and 32.7% lower, respectively, in MFCs with substrate pre-treatment compared to those without pre-treatment. In contrast, other device characteristics were higher in MFCs with pre-treatment than in those without. We also found that fluid and solid substrates had strong effects on the majority of MFC device characteristics, with the exception of internal resistance. For start-up time, reaction chamber volume, external resistance, and pH, MFCs with fluid substrates showed values that were 6.6–11.9% and 37.4–50.1% higher than those with liquid and solid substrates, respectively. Regarding electrode surface area, reaction duration, and temperature, MFCs with solid substrates had 23–25% higher values than those with fluid and liquid substrates.

The results of the t-test and one-way ANOVA showed that the power generation performance of MFCs was significantly affected by reaction chamber type, substrate operating phase, and pre-treatment (Figure 4). The power density of MFCs without substrate pre-treatment was 44.02 ± 4.54 mW/m² and 2.33 ± 1.79 W/m³, which were 10.5 and 15.5% higher, respectively, than those of MFCs with pre-treatment. By contrast, the current density of MFCs without substrate pre-treatment was 133.96 ± 5.94 mA/m² and 10.79 ± 11.47 A/m³, which were 13.1 and 12.3% lower, respectively, than in MFCs with pre-treatment. Moreover, voltage followed the same trend, decreasing by 4.9% after pre-treatment. Dual-chamber MFCs showed ~15% higher values in the majority of electricity generation indices compared to single-chamber MFCs, except for voltage and power. The voltage in dual-chamber MFCs was only 3% higher than that in single-chamber MFCs, whereas the power output was 80% higher. The power density and voltage of MFCs using solid substrates were 77.73 ± 6.28 mW/m² and 235.99 ± 1.53 mV, respectively, which were 10 and 7% higher than those using liquid and fluid substrates. However, volumetric power density showed no significant differences among substrate operating phases. The coulombic efficiency of MFCs with liquid substrates was 25.92 ± 2.79%, which was nearly half that of solid substrates. Current density was highest (108.73 mA/m²) for liquid substrates, while volumetric current density was highest (643.19 A/m³) for solid substrates.

The majority of continuous variables related to device configurations and reaction conditions had significant impacts on power generation performance (Figures 5–7, Supplementary Table S3). Among these, the most influential factors were cathode chamber volume and cathode surface area. Cathode surface area significantly affected all power generation performance indices, except voltage, with R² values ranging from 0.01 to 0.257. It had the lowest explanatory power for power output but the highest for coulombic efficiency (R² = 0.257). Furthermore, the cathode surface area exhibited the highest positive correlation with power (R = 0.313) and the highest negative correlation with coulombic efficiency (R = −0.529). Another important device configuration factor was cathode chamber volume, which showed the lowest explanatory power for voltage (R² = 0.01) but explained 43.1% of the variance in coulombic efficiency. Similar to the cathode surface area, cathode chamber volume showed the strongest positive correlation with power (R = 0.493) and the strongest negative correlation with coulombic efficiency (R = −0.653). Other noteworthy factors included pH and reaction duration. Reaction duration significantly influenced all power generation performance indices except for coulombic efficiency, with R² values ranging from 0.01 to 0.59. It had the weakest explanatory power for voltage (1.0%) and the strongest for power (59.5%). Specifically, our result showed that power density was negatively correlated with reaction duration between 0 and 32.11 h and positively correlated between 32.11 and 1,095.63 h (R² = 0.21). Moreover, pH significantly affected all indices except volumetric current density (A/m³), with R² values ranging from 0.01 to 0.59. It was negatively correlated with power density (R² = 0.116) but showed positive effects on the other power generation performance indices, explaining 85.1% of the variance in coulombic efficiency.

The multiple regression analysis showed that among all device configurations, the most important factors were cathode surface area and cathode chamber volume (Table 2). Cathode chamber volume significantly affected all power generation performance indices, with the highest explanatory power for volumetric power density (R² = 0.813) and the lowest for current density (R² = 0.128). Regarding the cathode material area, it affected power density, current density, and coulombic efficiency, with R² values ranging from 1.0 to 33.4%. Among all electricity generation performance indices, device configurations showed the best goodness of fit for coulombic efficiency (R² = 0.891). For reaction conditions, the most important factors were reaction duration and external resistance. Reaction duration had the weakest explanatory power for voltage (R² = 0.014) but the strongest for volumetric current density (R² = 0.759). External resistance was significantly correlated with all electricity generation performance indices except power density, with a maximum R² value of 0.742. Among all electricity generation performance indices, reaction conditions showed the best goodness of fit for volumetric current density (R² = 0.785). When considering both device configurations and reaction conditions together, the best-fitting model was for coulombic efficiency (R² = 0.941). Notably, both device configurations and reaction conditions explained <5% of the variance in voltage. In addition, continuous and categorical variables showed complex interaction effects on electricity generation performance (Supplementary Table S4). For the majority of electricity generation performance indices, if the main effects were significant, the interaction effects were also significant. However, even if the main effects were not significant, the interaction effects may still be significant. For example, the effect of the substrate operating phase on current density was not significant on its own; however, when the cathode surface area was considered, the interaction effect became significant.

All power generation performance indices examined in our study differed from those reported in other meta-analyses or review papers. The relatively lower power density in our study may be due to our inclusion of data from the entire stage of the polarization curve for each experiment, while other studies collected only the highest values. The relatively higher coulombic efficiency observed in our study compared to other meta-analyses or review papers may be due to the imbalanced sample size between dual-chamber and single-chamber MFCs. Our dataset included 6,457 cases of dual-chamber MFCs, which was ~2.1 times the number of single-chamber cases. Since the coulombic efficiency of dual-chamber MFCs can be 73% higher than those of single-chamber systems, as supported by our results and other studies (Khan et al., 2018), this imbalance may have contributed to the elevated efficiency. Additional factors contributing to differences between our study and others include variations in sampling size, substrate characteristics (e.g., some studies focused exclusively on food waste; Zafar et al., 2022; Dowdy et al., 2018), and device configurations (e.g., some studies concentrated on anode modification data; Agrahari et al., 2022). Although our study was limited to Chinese papers, which may have caused some bias, our study had clear advantages. Specifically, we compiled a dataset of 11,806 cases, considerably larger than those in other studies (which ranged from dozens to several thousand). In addition, our dataset included a relatively large number of variables, providing a more comprehensive understanding of MFC characteristics and enabling the identification of key driving factors for power generation performance. However, both our study and other review papers reported much lower values for power generation performance metrics—namely voltage, power, power density, current density, and coulombic efficiency—compared to commonly used batteries (e.g., lithium and lead–zinc batteries). This highlights the ongoing challenge of improving the generally low power generation performance of MFCs, which remains a major challenge for future studies (Gupta et al., 2023; Logan and Rabaey, 2012).

We found that all MFC device configuration variables significantly correlated with at least one power generation performance indicator, with the strength of these relationships ranging from 0.5 to 89.1%. Our results suggest that the effects of device configurations on power generation performance are context-dependent. Among these variables, cathode chamber volume, cathode surface area, and anode surface area emerged as the strongest predictors, as indicated by our linear and multiple regression models. More than 80% of the variances in coulombic efficiency and volumetric power density could be explained by cathode chamber volume and cathode surface area. This result aligns with previous studies, which have also reported positive relationships between power density and both cathode chamber volume and cathode surface area (Opoku et al., 2022; Logan et al., 2007). For example, Cheng and Logan (2011) found that cathode surface area was the most critical factor influencing power density: doubling the cathode surface area increased power by 62% while doubling the anode surface area led to only a 12% increase. A larger cathode surface area provides more room for microbial colonization, thereby enhancing power generation performance (Wang et al., 2024). Additionally, a larger cathode area can increase the rate of the oxygen reduction reaction, a crucial process in MFCs. With greater oxygen availability at the cathode, the overall efficiency of electron transfer improves, leading to a significant increase in power output (Elmekawy et al., 2017). We also observed significant differences in power generation performance between single- and dual-chamber MFCs. Dual-chamber MFCs showed higher values for voltage, power density, current density, and coulombic efficiency, while single-chamber MFCs had higher power output. Our results are in line with those of many other studies (Zafar et al., 2022; Lee et al., 2016). Although dual-chamber MFCs were more frequently studied than single-chamber MFCs, this does not necessarily imply superior performance across all indices. Other influencing factors—such as organic matter loading rate, substrate pre-treatment, and substrate type—may cause extremely high or low power density (Zafar et al., 2022). Notably, we found that cathode chamber volume may exert a stronger effect on power generation performance than anode chamber volume. This implies that the common practice of designing MFCs with equal anode and cathode chamber volumes may need reconsideration to optimize power output (Qian et al., 2009).

Electrical production performance was also affected by chamber type. MFCs with dual chambers exhibited higher voltage, power density, current density, and coulombic efficiency than single-chamber MFCs, partially confirming previous studies (Zafar et al., 2022; Jia et al., 2008). The relatively high coulombic efficiency in dual-chamber MFCs compared to single-chamber MFCs is likely due to the higher levels of dissolved oxygen in single-chamber systems, which may adversely affect microbial activities (Nimje et al., 2012), as well as shifts in microbial communities resulting from prolonged oxygen diffusion in single-chamber MFCs (Flimban et al., 2019). Although we found lower internal resistance in single-chamber MFCs, which should have led to higher power density, we still observed lower power density compared to dual-chamber systems, suggesting that other contributing factors are involved. For example, Lee et al. (2016) found that the abundance of Proteobacteria was 2–3 times higher in dual-chamber MFCs than in single-chamber MFCs and that the dominant archaeal species differed between the two types. Regarding power, the relatively small dataset (177 cases) may have introduced bias, which should be taken into consideration.

Our results indicated that ambient temperature, external resistance, reaction duration, and pH showed stronger effects on the power generation performance of MFCs than other reaction conditions. These results are consistent with previous studies (Pasupuleti et al., 2015; Zhao et al., 2014; Zhang et al., 2010). Ambient temperature significantly affected the power production performance of MFCs. We found that the optimal ambient temperature for maximum power generation performance was 28°C, which falls in the lower range (25–35°C) reported in other studies (Uddin et al., 2021). The majority of studies have shown positive effects of temperature on MFC performance, while performance declines when the temperature exceeds a certain threshold (Hiegemann et al., 2016; Li et al., 2013). This is because microbial activity generally increases with temperature, and many chemical reactions involved in MFC operation also perform optimally at 30–35°C (Zhao et al., 2014). Changes in temperature not only significantly affect the oxidation rates of organic matters but also influence the species diversity (richness and evenness) of anode biofilms, which, in turn, affects metabolic activity, biofilm accumulation, and extracellular electron transfer—all of which impact MFC power generation (Mei et al., 2017). In addition, changes in temperature can affect internal resistance and pH (Pasupuleti et al., 2015). For example, Li et al. (2013) found that internal resistance was approximately 29 Ω at 37°C, but it increased by 62%, 303%, and 48% at 30°C, 10°C, and 43°C, respectively. Other reaction conditions—such as external resistance, internal resistance, and reaction duration—also affected the power generation performance of MFCs. According to Grondin et al. (2012), MFCs are typically operated with a constant exterior resistance, but their internal resistance varies over time. This mismatch reduces power generation efficiency. The highest power output is generally achieved when external resistance matches the total internal resistance (Lyon et al., 2010). External resistance also determines the quality of energy available for the growth and maintenance of microorganisms: bacteria obtain sufficient energy only at lower external resistances, which are associated with lower anode potentials and higher electron fluxes (Aelterman et al., 2008). In addition, our results suggested that a neutral pH condition yielded better performance across the majority of indices compared to acidic or alkaline conditions. This finding is consistent with many previous studies (Tremouli et al., 2017), even though some researchers have reported better power generation performance under acidic or alkaline conditions than at neutral pH (Dekker et al., 2009; He et al., 2008). This is likely because a neutral pH is optimal for the majority of microbes, whereas microbial activities can be inhibited—or even halted—at pH levels that are too high or too low (Puig et al., 2010).

Substrate characteristics were also key factors affecting the power generation performance of MFCs. We found that substrates with pre-treatment significantly improved power density, voltage, and power, which were in line with other studies (He et al., 2025; Li et al., 2024).

By increasing the biodegradability of substrates, pre-treatment can boost the microbial utilization efficiency of the inoculum. Substrates can generally be pre-treated using physical, chemical, or biological, or combined (two or more pre-treatments) methods to modify the chemical structure of organic matter and support electricity-producing bacteria (He et al., 2025; Oh et al., 2014). Pre-treatment increases the biodegradability of substrates, making it easier for them to break down organic matter and release more organic compounds, thereby improving power density, voltage, and power output (Ray and Ghangrekar, 2015). Pre-treatments can also enhance core metabolic pathways—including those involved in energy, carbohydrate, nucleotide, lipid, and amino acid metabolism—and change the metabolite components of biofilms, thereby improving the power generation performance of MFC (He et al., 2025). Furthermore, our results suggested that biological pre-treatment yielded the highest power density, while physical and chemical pre-treatment were more effective at improving voltage and coulombic efficiency. However, this finding contrasts with a recent meta-analysis that found no significant differences in power generation performance among pre-treatment methods (Zafar et al., 2022). This discrepancy may be explained by the gentler nature of biological pre-treatment, in contrast to physical and chemical approaches, which often involve crushing, high temperatures, and the use of chemical catalysts or additives (Oh et al., 2014; Cao et al., 2018). These harsher treatments can damage microbial structures, reducing the effectiveness of power density improvement. To achieve higher power density, biological pre-treatments often include microbial inoculation, aerobic fermentation, or anaerobic fermentation to prevent the formation of inhibitory substances in the MFCs. For example, Rajesh et al. (2015) found that methane production reduced MFC power output by depleting available substrates. To address this, he inoculated the marine diatom Chaetoceros using a biopreparation method, which reduced methanogenic activity by 60%. This led to an increase in power density and overall performance. The non-significant effect of substrate pre-treatment on coulombic efficiency may be due to the fact that the majority of substrates in our study were simple, and their coulombic efficiency may not be significantly impacted by pre-treatment. Current densities in the unpretreated group were consistently higher than those in the pre-treated group. This could be because biological pre-treatment overly degraded the substrates, producing short-chain fatty acids such as acetic acid, propionic acid, or phenolics that inhibit the proper functioning of the electron transport chain in electroactive bacteria. As a result, the treatment's current density may have been somewhat lower than before treatment (Hassan et al., 2012). Chemical pre-treatment may also leave behind residual substances—such as H⁺, Cl⁻, free radicals, or remnants of acid, alkali, or oxidant pre-treatment—that can improve power generation performance at optimal concentrations but may exhibit toxic effects at higher doses. These residues can directly harm electroactive bacteria or damage the structure of the biofilm (He et al., 2025).

Additionally, the use of carbon felt as the electrode material, which is prone to pore blockage, can reduce biofilm binding sites and lower current density. This issue is further exacerbated by excessive grinding, high temperature, high pressure, or other treatments that result in substrate particles that are too small. Therefore, it is important to carefully consider whether pre-treatment methods may negatively impact the substrate or the entire MFC system, potentially leading to reduced current density. In summary, pre-treatment generally enhances power generation performance, and biological pre-treatment showed greater improvements compared to other pre-treatment methods.

In addition, our results indicated that solid substrates produced higher power, voltage, and current densities compared to liquid and fluid substrates, while liquid substrates exhibited higher coulombic efficiencies. These findings are consistent with previous studies that found that solid substrates had higher electrical performance than liquid and fluid substrates (Touch et al., 2014; Zhao et al., 2017), although they contradict others that suggested that solid substrates limited mass transfer due to insufficient physical transport rates of reactants at the electrode interface or within the reaction system. This limitation may result in a lower electrochemical reaction rate than the theoretical maximum, thereby restricting the electrical performance of MFCs. To improve the stability of power production performance, complex substrates are typically pre-treated—physically, chemically, or biologically—to reduce particle size, break down complex polysaccharides into monosaccharides, and inhibit methane production. This supports the earlier discussion on the effectiveness of pre-treatment methods. The majority of studies have used liquid and fluid substrates (e.g., wastewater and sludge), while relatively few have explored the use of solid substrates. Therefore, future studies may need to further investigate whether solid substrates consistently perform better for power generation than liquid and fluid substrates.