Production of microalgae in wastewater and brackish waters: kinetic, lipid content, bioremediation and cost analysis studies

The open cultivation of microalgae has become increasingly relevant, as there is a wide range of applications for these microorganisms, ranging from biofuel production to water purification. External cultivation environments can be ponds, tanks or cultivation ponds, unlike closed environments, which include controlled systems such as photobioreactors. The main source of energy in the open method is sunlight, used to carry out photosynthesis. The advantages of open microalgae cultivation include the lower effective cost, as less initial investment is required compared to closed systems, making it more accessible. Another aspect is the use of local resources such as sunlight and atmospheric CO₂, which are abundant and free. It is also possible to highlight the possibility of easily increasing the scale of cultivation, adapting to larger areas as necessary. Another important aspect is the wide applicability, with biomass produced in open systems being able to be used in various applications, such as biofuels, food production, environmental remediation and pharmaceutical products.

It is also important to note that the open cultivation of microalgae has some disadvantages, such as exposure to the external environment, increasing the risk of contamination by unwanted microorganisms. Furthermore, climatic conditions, such as variations in temperature and light, can affect the productivity of biomass production. Another factor is the difficulty in maintaining adequate concentrations of nutrients, which is an essential factor for successful cultivation. The open cultivation of microalgae continues to be an area of active research and ongoing study, based on the development of more efficient cultivation techniques and the selection of more productive species. With the growing interest in sustainability and the search for renewable sources of energy and resources, the cultivation of microalgae in an open environment has an important role to play (Chisti 2007; Gouveia & Oliveira 2009; Borowitzka & Moheimani 2013; Richmond & Hu 2013).

The production of microalgae biomass in brackish environments has aroused considerable interest due to its high potential to increase lipid production, which is of great importance, as it can be used in the production of biofuels, such as biodiesel, and in high-value products, aggregate, such as oils rich in omega-3, applicable in the food and pharmaceutical industries. Many species of microalgae are naturally found in marine or saline habitats, making them well adapted to growing in brackish environments. These media encourage microalgae to accumulate lipids as a way of storing energy in response to environmental stress. It is an intracellular defense mechanism that leads to an increase in lipid content under salt stress. Several species are known to be oilseed and can be successfully cultivated in brackish environments. Examples include Chlorella vulgaris, Nannochloropsis sp., and Dunaliella salina. These species have been widely studied due to their ability to produce high-quality lipids under saline conditions. The accumulated lipids can be extracted and converted into biodiesel through transesterification processes, which can meet part of the current energy demand. Despite the advantages mentioned, challenges are still faced in terms of controlling contamination and optimizing cultivation conditions (Hu et al. 2008; Converti et al. 2009; Ben Amor et al. 2015).

In addition to the use of brackish media, it is also possible to use domestic effluents as a source of nutrients, representing an environmentally sustainable approach to treating this waste and generating biomass, with added value. Domestic effluents contain a variety of nutrients, including nitrogen, phosphorus and organic carbon, which are essential for microalgal growth. These high levels of nutrients make domestic effluents an attractive and economically viable source for the cultivation of microalgae, at the same time contributing to the reduction of water pollution levels, as microalgae consume the nitrogen and phosphorus present in the effluent, transforming it into into biomass and resulting in a treated effluent, with reduced levels of nutrients, which can be more easily discarded into the environment. The field of study of microalgae biomass production using domestic effluents as a source of nutrients still presents challenges, such as the selection of microalgae species, the control of cultivation conditions and the harvesting of biomass (Converti et al. 2009; Mata et al. 2010).

Within the context of production, the use of computational tools in predicting kinetic models of microalgae production on a pilot scale has become an important tool, aiming to optimize cultivation processes and maximize production efficiency. Kinetic modeling allows us to better understand the dynamics of microalgae growth in response to different cultivation conditions to predict cell concentration over cultivation time and identify critical factors that affect their performance. Furthermore, it provides crucial information about growth rates, nutrient consumption, biomass production and accumulation of compounds of interest, such as lipids and pigments. Using simulation software allows researchers to create mathematical models that accurately represent the complex interactions between cultivation variables such as temperature, lighting, nutrient concentrations and cell growth rate.

Kinetic modeling helps predict the responses of microalgae on a pilot scale. This becomes an important tool for scaling and optimizing large-scale cultivation systems. Such tools allow extrapolating the results obtained on a pilot scale to large-scale production, saving time and resources. Through sensitivity analysis and tuning model parameters with experimental data, researchers can find the optimal conditions that maximize production. In summary, the use of computational tools to predict kinetic models of microalgae production on a pilot scale is essential to advance the research and development of this promising technology. It not only assists in understanding the biological processes involved, but also facilitates the optimization of cultivation conditions, contributing to the efficient and sustainable production of microalgae for a variety of applications, from biofuels to food and pharmaceuticals (Wang & Hu 2017; Koller et al. 2014).

Among the factors that are responsible for a large part of production costs, the process of separating the biomass produced is one of the main processes responsible for the increase in costs. Ceramic membranes are known for their ability to retain microalgae due to their tiny, uniform pores. This allows for the efficient separation of microorganisms of varying sizes, making them particularly useful in microalgae cultivation and biomass recovery applications. Ceramic membranes are highly durable and chemical resistant, which is critical in microalgae separation applications. They can be exposed to corrosive environments and cleaning agents without significant degradation, which extends the life of the membranes. In water and wastewater treatment systems, the use of ceramic membranes is crucial for removing microalgae and other potentially harmful microorganisms. This helps prevent algae blooms and reduces the formation of unwanted compounds such as cyanobacteria and their toxins. Ceramic membranes play an important role in the production of biofuels from microalgae. They allow the efficient separation of microalgae from the culture, facilitating the extraction of valuable lipids for the production of biodiesel or other biofuels (Dizge et al. 2019; Zhu et al. 2020; Chi et al. 2021).

In this scenario, the present work proposes the production of biomass from different species of microalgae, using water from brackish wells and using septic tank effluents as a source of nutrients, in different proportions. The use of these means brings environmental benefits, from the reuse of septic tank effluents, observing a reduction in the polluting load, as well as the levels of ammonia and other nutrients and the number of Escherichia coli bacteria, the main indicator of fecal contamination, collaborating in the decontamination of water bodies.

Still aiming to reduce production costs, ceramic membranes are used in the process of separating biomass from the cultivation medium, considering that conventional processes such as centrifugation, coagulation, flocculation, filtration, among others, are responsible for a large part of the production costs, which has often made large-scale production of microalgae biomass for biodiesel production unfeasible. Ceramic membranes were produced with low-cost material and can be reused, which further contributes to reducing production costs.

It was identified which species, among those studied, have the best lipid levels versus biomass production, and the potential for biodiesel production from microalgae biomass was verified. This research has the distinction of, through a cost analysis, bringing the possibility and viability of implementing large-scale biomass production systems, using domestic sewage from small- and medium-sized cities and wells in the region as a means of cultivation, aiming at the production of biodiesel, which can promote the economic and social development of the region. In addition to the above, computational tools are used to study kinetic models for cell growth, and comparisons are made with experimental results.

The computational tools aim to expand the study of the kinetic mechanisms of microalgae. Studies of the kinetics of microorganisms in aqueous media cannot be summarized in simple equations and models, requiring the separation of multiple stages and equations corresponding to lag, log, stationary and death phases. Using machine learning, the study aims to determine the possibility of creating an artificial neural network (ANN) to determine the growth kinetics of microalgae strains. The ANN must produce a method that comprises all stages of the microalgal growth phases, without splitting into multiple equations for the different growth stages.

The creation of a tool for the accurate prediction of microbial concentration in a medium allows the simplification of scale-up processes and the production of continuous flow microalgal growth reactors. Therefore, this work included the use of microalgae concentrations in solution to investigate the possibility of using ANNs to determine the growth kinetics of strains of Chlorella sp., Scenedesmus acuminatus and Nannochloropsis sp. The studies carried out in this context aimed to verify the ability of ANNs to adapt to microalgal growth behavior, and adequate representation of the operational phases of microorganism kinetics.

In the case of open reactors, control of cultivation parameters is limited and subject to change due to the impact of external climatic effects such as changes in temperature, rain and lighting. Therefore, the application of such tools is more complex, but can be used in specific reactors and fixed locations. This type of study is applied to validate the process in open reactors, and is not applied to other reactors or different cultivation conditions. Validation of the process itself, however, can be applied to the creation and use of similar processes in external reactors for large-scale application. The nature of microbiological processes, which use external factors such as light and temperature variations, favors the use of open reactors in this type of process. In this case, the use of tools to generate a robust kinetic model allows the use of large-scale bioreactors and the construction of controllers for better process management.

ANNs have been commonly used for the control and design of reactors that use bioprocesses, such as Najafzadeh et al. (2021), which uses ANNs to determine water quality in the Karun River, Iran. This type of use has proven to be a practical way to observe phenomena that have natural randomness of samples as parameters in continuous water flows. In this way, the study of parameters with high variability and the influence of external factors can be simplified with the use of flexible computing, obtaining high precision for measurements (Najafzadeh & Zeinolabedini 2019).

With the use of ANN, therefore, this paper aims to determine tools for measuring the growth kinetics of microalgae for open reactors, with the intention of simplifying the way in which microalgal growth is observed and reducing errors such as behavior at phase boundaries and excessive simplifications of the equations.

The study was developed at the Universidade Federal de Campina Grande (UFCG), in Campina Grande, Paraíba, Brazil. The species were isolated from the region itself and maintained in an environment with controlled temperature and light. The species Chlorella sp., of the genus Chlorophyta, was cultivated, which was isolated from the region itself, having a spherical shape and a diameter varying between 0 and 10 μm. The species S. acuminatus, and the microalgae of the genus Nannochloropsis sp., which has six known species, native to saline and brackish environments, were also studied. The strains of the studied species were replicated in synthetic culture medium, produced in the laboratory. Two tanks were adapted for the production of biomass of the studied species, with a cultivation volume of 1,000 L in each tank. The system was maintained under constant agitation using stirring paddles, to homogenize the medium and capture light evenly by the cells. Over time, open crops were exposed to environmental temperature variations, as well as subjected to natural light.

For the cultures, the media were prepared for subsequent inoculation in the tanks under constant agitation, these media being composed of: brackish water from a well, located close to the tank installations and raw sewage water from septic tanks located on the campus itself. The effluent/well water ratio used was 5% by volume for Chlorella sp. and 50% for S. acuminatus and Nannochloropsis sp., according to a previous study carried out by Guimarães (2020). Cultivations were carried out in duplicate. Therefore, in the biomass production batch of the species in question, the same proportions were used in both tanks. After mixing the two media, they were characterized in terms of concentration of nutrients NH₃, ⁠, and total phosphorus, as well as quantification of the number of colonies of E. coli bacteria, chemical oxygen demand (COD) and biochemical oxygen demand (BOD). All analyses followed methodologies described in the Standard Methods for the Examination of Water and Wastewater (APHA 2023).

After inoculating the species, cell growth was monitored by counting the number of cells daily. Once the stationary growth phase was reached, the biomass and cultivation medium were separated. A separation system was set up with ceramic membranes, produced in the laboratory, using alumina and clay. Membranes with apparent porosities were used, amounting to 51.98 and 42.79%. They had an average useful length of 15.77 ± 0.86 cm, with a useful area of 0.012 ± 0.001 m². The separation system operated in ‘dead end’, not generating an output stream of the concentrate, which accumulated in the containers that contained the membranes. The separation system was designed specifically for open cultivations on a pilot scale, therefore, for the volume of cultivations, the system configuration was the most appropriate, seeking to optimize biomass recovery.

In arranging the separation system, eight ceramic membranes were used, arranged in two blocks of four operating in parallel. In each block, the first two membranes, with greater apparent porosity, received the cultures, which were pumped from the tanks and operated in series with the other two membranes with lower porosity, which were fed with the permeate from the first. Tests were carried out to obtain the best operating conditions, in order to maximize separation efficiency, in terms of permeate flow rate, volume of concentrate obtained versus feed pressure, which in all cultures was 3.3 bar, this being the one that showed the best efficiency in obtaining the concentrate. Once the process was complete, the system was dismantled and the ceramic membranes were chemically cleaned to unclog the pores, using a 1.0 N sodium hydroxide solution and an ultrasonic bath.

Once the separation occurred, the quantification of nutrients and pollutant load carried out at the beginning of the cultures was repeated, as well as the number of colonies of E. coli bacteria present in the medium after cultivation. In this way, factors such as percentages of nutrient removal were analyzed, as well as decontamination levels, through the quantification of this bacteria, the main indicator of fecal contamination in water bodies, in addition to the percentages of removal of the polluting load, through analyses of COD and BOD. Quantifications of lipid content and dry biomass production per liter of cultivation were also made, through drying in an oven at 45 °C. Lipid content analyses followed the methodology recommended by Folch et al. (1956) modified for microalgae.

Based on literature research, an analysis of the costs involved in the production of microalgae biomass on a large scale and subsequent production of biodiesel was carried out. Price estimates were made based on literature and market research. Cost calculations were based on large-scale production, with annual production of 10 million gallons, in accordance with Davis Aden & Pienkos (2011). Azeredo (2012) calculated the costs of construction and implementation of open raceway-type cultivation tanks for a built area of 200 hectares, which were taken as the basis for the present study. It is important to highlight that in the present work the stages of strain selection, cultivation, separation, pre-treatment, obtaining dry biomass and extraction of lipids for quantification were carried out. The subsequent steps are based on the study by Davis Aden & Pienkos (2011) which shows the biodiesel production process.

Based on the data obtained in the experimental stages, regarding growth kinetics, the MATLAB R2021a software was used to train the neural networks to determine a growth kinetic model. The software was chosen due to the presence of the artificial intelligence Toolbox, making the process more simplified. The choice of the individual characteristics of each neural network was made by trial and error, using the training function provided by the software and various numbers of neurons per intelligence. The results obtained were determined based on the coefficient of determination (R²). Separate neural networks were trained for each curve due to the high experimental error found in the previous stages, as the lack of control over parameters such as temperature, luminosity, pH of the environment, occurrence of rain, among others, interfered with the final result.

The ANNs used received information for training on the microalgae concentrations observed during the analyses carried out throughout the experiments and the time elapsed between each measurement. All information on concentrations obtained and time was used in conjunction with the MATLAB Neural Network Training Toolbox to determine the training. Regarding the determination of the number of neurons used in each neural network, the values were determined using a trial-and-error method for each of the strains. All experiments performed produced a microalgae growth kinetics curve with machine learning, thus producing six ANNs, two for each strain of microalgae.

The training of the neural networks was carried out to solve the calculation of expected concentration in an aqueous medium of microalgae mass. The trained neural network receives input values for the initial concentration of microalgae in the system, total nitrogen concentration in the aqueous medium and time elapsed since the beginning of the experiment. The output of the neural network produces the value estimated by the ANN for the current concentration of microalgae. The precision value of the output values of the neural network was defined by the coefficient of determination, R² of the obtained results and the experimental results.

Network training used the standard Matlab tool for data optimization, using the Levenberg-Marquardt algorithm for ANN optimization. Promising results for training neural networks in water processes were obtained by Saberi-Movahed et al. (2020) using the data processing group method (GMDH), Najafzadeh et al. (2019) using gene expression programming (GEP), evolutionary polynomial regression (EPR) and model tree (MT); however, these models, despite being more sophisticated, depend on a large volume of data to produce more accurate results. Due to the low number of data used, the simplest method, Levenberg-Marquardt, and trial-and-error repetitions were therefore chosen for ANN optimization.

For S. acuminatus there was also similarity between the curves, with the cultivation period being 7 days for both tanks. The maximum number of cells was reached on the fifth day of cultivation, being slightly higher in tank 1 than in 2. Regarding Nannochloropsis sp., differences were noticed between the two curves in the two culture tanks, which may be due to the fact that these were carried out on different days where there were different weather conditions. The cultivation time was also 7 days in both tanks. It is important to highlight that the batch of biomass production in the second occurred on rainy days, where the luminosity rate was low, a fact that directly affected cell growth. This explains the lower cell concentrations in tank 2. Such results prove the dependence of cell growth on external factors, mainly light rate and temperature.

By analyzing the growth curves, it was possible to observe the exponential growth phase, or log phase, and, from a linear regression, calculate the maximum growth speed (μ_m), expressed in day⁻¹, as well as the generation time (t_g), given in days. The results obtained are found in Table 1. For Chlorella sp., the maximum growth speed was higher in tank 1, which can be justified by the fact that, although the cultures were carried out simultaneously, when the effluent was collected, part of the suspended material settled in the collection container, leaving the water used more loaded with nutrients and polluting load, a fact that may have affected the speed of microalgal growth. Therefore, in tank 2, both the high concentration of nutrients and the high content of organic matter may have negatively influenced the growth kinetics. These results show the importance of the kinetic study of the species that is cultivated, mainly on a large scale, depending on the concentrations of nutrients available in the cultivation medium, making it necessary to verify the optimal concentration, which will maximize biomass production, as well as the its growth kinetics.

Table 1

Growth kinetics, nutrient concentrations, pollution load and initial and final Escherichia coli

.	Chlorella sp. .				Scenedesmus acuminatus .
	Tank 1 .		Tank 2 .		Tank 1 .		Tank 2 .
	Initial .	Final .	Initial .	Final .	Initial .	Final .	Initial .	Final .
(mg·L−1)	0.55	0.76	0.45	4.9	1.77	0.33	0.1	28.62
(mg·L−1)	0.235	0.217	0.157	1.59	0.779	0.308	0.021	19.598
N–NH3 (mg·L−1)	11.12	2.97	19.64	4.3	82.66	0.69	120.58	9.79
P-Total (mg·L−1)	2.7	8.9	0.4	23.0	8.0	25.1	5.4	17.1
DQO (mg of O2.L − 1)	119.66	264.96	188.03	188.03	230.77	17.09	174.60	238.09
DBO5 (mg of O2.L-1)	45.38	18.77	127.84	49.39	49.1	5.36	6.9	1.36
Escherichia coli (CFU.mL−1)	1.1 × 102	6.2 × 100	5.3 × 103	8.3 × 100	5.5 × 104	4.0 × 103	1.0 × 103	1.0 × 102
μm (day−1)	1.0099	0.6393	1.0938	0.9268
tg (day)	0.69	1.08	0.63	0.75
.	Nannochloropsis sp. .				.	.	.	.
	Tank 1 .		Tank 2 .		.	.	.	.
	Initial .	Final .	Initial .	Final .	.	.	.	.
(mg·L−1)	1.08	4.85	0.08	5.21
(mg·L−1)	0.52	1.597	0.018	0.651
N–NH3 (mg·L−1)	73.52	2.12	72.15	57.31
P-Total (mg·L−1)	4.7	9.7	3.4	12.8
DQO (mg of O2.L-1)	247.87	51.28	230.16	214.29
DBO5 (mg of O2.L-1)	47.9	21.6	7.65	2.06
Escherichia coli (CFU.mL−1)	5.0 × 104	3.0 × 103	1.0 × 103	1.0 × 102
μm (day−1)	1.0109		0.4823
tg (day)	0.69		1.44

.	Chlorella sp. .				Scenedesmus acuminatus .
	Tank 1 .		Tank 2 .		Tank 1 .		Tank 2 .
	Initial .	Final .	Initial .	Final .	Initial .	Final .	Initial .	Final .
(mg·L−1)	0.55	0.76	0.45	4.9	1.77	0.33	0.1	28.62
(mg·L−1)	0.235	0.217	0.157	1.59	0.779	0.308	0.021	19.598
N–NH3 (mg·L−1)	11.12	2.97	19.64	4.3	82.66	0.69	120.58	9.79
P-Total (mg·L−1)	2.7	8.9	0.4	23.0	8.0	25.1	5.4	17.1
DQO (mg of O2.L − 1)	119.66	264.96	188.03	188.03	230.77	17.09	174.60	238.09
DBO5 (mg of O2.L-1)	45.38	18.77	127.84	49.39	49.1	5.36	6.9	1.36
Escherichia coli (CFU.mL−1)	1.1 × 102	6.2 × 100	5.3 × 103	8.3 × 100	5.5 × 104	4.0 × 103	1.0 × 103	1.0 × 102
μm (day−1)	1.0099	0.6393	1.0938	0.9268
tg (day)	0.69	1.08	0.63	0.75
.	Nannochloropsis sp. .				.	.	.	.
	Tank 1 .		Tank 2 .		.	.	.	.
	Initial .	Final .	Initial .	Final .	.	.	.	.
(mg·L−1)	1.08	4.85	0.08	5.21
(mg·L−1)	0.52	1.597	0.018	0.651
N–NH3 (mg·L−1)	73.52	2.12	72.15	57.31
P-Total (mg·L−1)	4.7	9.7	3.4	12.8
DQO (mg of O2.L-1)	247.87	51.28	230.16	214.29
DBO5 (mg of O2.L-1)	47.9	21.6	7.65	2.06
Escherichia coli (CFU.mL−1)	5.0 × 104	3.0 × 103	1.0 × 103	1.0 × 102
μm (day−1)	1.0109		0.4823
tg (day)	0.69		1.44

View Large

For S. acuminatus, the maximum growth speed and generation time remained close between the cultures in tanks 1 and 2. This occurred despite the fact that the two cultures were carried out at different times, due to the unavailability of sufficient effluents to carry out the cultivation of crops simultaneously. In cultivations of Nannochloropsis sp. cell growth and kinetics were much lower in tank 2, a fact that may have occurred due to the cultivation conditions and the influence of external factors that were out of control in open cultures. Liyanaarachchi et al. (2021) show that the biochemical composition of the biomass produced is a function of the cultivation conditions, that is, the parameters that directly affect growth. When these conditions promote rapid growth, low levels of intracellular compounds such as lipids, carbohydrates and carotenoids are often observed.

It is important to note that the growth kinetics of microalgae in open cultivation systems often differ considerably from theoretical kinetic models. While theoretical models often assume ideal and homogeneous conditions, open farming systems face a number of challenges, such as variations in temperature, light, nutrients and competition with other species. These differences can result in growth rates that do not follow predictable patterns, making the direct application of theoretical models difficult. The implications of these disparities are significant, especially in industrial applications, where optimizing microalgae growth is essential for the efficient production of biofuels, food or pharmaceuticals. Therefore, understanding and incorporating the complexity of open cultivation systems is crucial to developing effective microalgae cultivation strategies.

Table 1 also presents the results of nutrient analyses, pollutant load (COD and BOD), concentration of E. coli colonies, given in colony forming units (CFUs) per mL present in the medium before and after cultivation in tanks 1 and 2. Based on these results, the potential for removing nutrients, reducing the pollution load and decontaminating the environment was analyzed. For ammonia, there was a reduction of 73.29 and 78.10% in Chlorella sp. in tanks 1 and 2, respectively.

In S. acuminatus the reduction was greater, 99.16 and 91.88%, and in Nannochloropsis sp. 97.11% in tank 1 and 20.56% in tank 2. This low percentage in the second tank may be related to the growth kinetics, which occurred with lower maximum speed and longer generation time. Given that the cultivation in this tank was carried out under rainy climatic conditions and low light, these external agents altered cell growth and consequently the absorption of nutrients.

These results prove that the reuse of domestic effluents as an exclusive source of nutrients for the production of microalgae biomass has the potential to achieve high levels of nutrient reduction, especially ammonia. This brings environmental benefits of great proportions, as the treatment of such effluents from the cultivation of microalgae can reduce the occurrence of eutrophication of water bodies, that is, the excessive increase in the concentration of nutrients in the environment, causing an imbalance in the ecosystem, causing great damage. Works such as those by Xin et al. (2010) confirm the results obtained in the present study, where the authors observed the effect of nutrient concentration, mainly nitrogen and phosphorus, on microalgal growth. The authors also achieved removal levels above 90%, confirming the efficiency of treating domestic effluents through the cultivation of microalgae. It is important to note that the percentage of removal, in addition to being a function of growth kinetics, is also a function of the cultivated species, as removal levels varied depending on the species studied.

Regarding nitrate and nitrite ions, there was an increase, mainly in tank 2. Krustok et al. (2016) state that organic nitrogen is absorbed by microalgae in different ways, but the preference for absorption is ammonia over nitrate and nitrite, and that in wastewater the concentrations of and are low, while the NH₃ and levels are high. Furthermore, Van Der Steen & Lens (2017) also highlight that in microalgae cultivation using such cultivation media, symbiosis occurs between microalgae and nitrifying bacteria, which absorb part of the ammonia and oxidize this component to nitrate and nitrite, thus increasing the final concentrations of such nutrients. Such increases were more pronounced in S. acuminatus cultivations in the second tank and in Nannochloropsis sp. in the first, indicating that the action of these bacteria was more effective in effluents with higher concentrations of ammonia. Despite the increase in nitrate and nitrite levels in the media, this fact does not preclude the production of biomass from the reuse of domestic effluents as a source of nutrients, as these can, after cultivation, be reused for new crops or for other purposes, such as irrigation of higher plants.

Thakur & Medhi (2019) emphasize that the ammonia present in wastewater is oxidized into nitrite and nitrate by the action of ammonia-oxidizing bacteria, aerobic autotrophs. In the aerobic oxidation process, ammonia is oxidized into nitrite by the action of ammonia monooxygenase enzymes and then the oxidation of nitrite to nitrate occurs through the enzyme nitrite oxidase. For this reason, the bacteria–microalgae interaction has been shown to be efficient in the cultivation of microalgae and cyanobacteria for bioenergetic purposes.

For total phosphorus, there was an increase in the final concentration in relation to the initial concentration in all crops. These results confirm those obtained by Delgadillo-Mirquez et al. (2016), who attribute these increases to cell ruptures that can occur throughout the cultivation period. The authors also show that this same fact was observed by Martínez et al. (2000), who observed the white color appearance of the cells. There were significant decreases in the biological pollutant load, a fact observed by quantifying BOD, reaching 89.1% removal in the cultivation of S. acuminatus. These results are very positive, as they indicate the reduction of organic matter present in domestic effluents due to the biomass produced, contributing powerfully to their treatment. Increases in COD can be attributed to increases in and concentrations at the end of cultivation.

As for the concentration of E. coli bacteria, there was a reduction of 2 logs in Chlorella sp. and 1 log for S. acuminatus and Nannochloropsis sp, in both tanks. Based on these results and taking into account that this bacteria is the main indicator of effluent contamination, it appears that the cultivation of microalgae, using domestic effluents as a source of nutrients, can reduce the levels of contamination of bodies of water: water receptors, to the extent that dumping can be done after crops or even these can be reused in new crops. Imagining the production of microalgae biomass on a large scale, large volumes of domestic effluents, rich in nutrients, would be reused, ratifying the environmental benefits.

Biomass separation tests using the filtration system with ceramic membranes proved to be efficient in terms of recovery, where, based on the color of the cultivation medium and concentrate obtained after separation, it was noticed that the biomass was almost completely retained in the concentrate. Difficulties were encountered in the filtration process regarding the clogging of the membrane pores over time, making it necessary to stop the process and disassemble the system for cleaning, using a 1 N NaOH solution to completely remove the organic matter from within the pores.

However, when analyzing the cost-benefit, and taking into account that the production of these membranes is a cheap process, where they are produced from alumina (Al₂O₃), a low-cost and easy-to-obtain material, they can be reused countless times. Sometimes, with only washing required to recover the initial separation efficiency, this fact does not preclude large-scale use. Therefore, it is possible to use this separation process, replacing conventional ones such as coagulation, centrifugation, flotation, among others, in large-scale production, potentially reducing production costs even further. It is important to highlight that, for large-scale production, the separation system must be sized depending on the intended production and volume of cultivation generated, requiring the study of parameters such as membrane surface area, filtration capacity and process conditions, such as such as operating pressure and efficiency in the filtration process.

Table 2 presents the percentages of intracellular lipids observed in the microalgae biomass produced in the present study, as well as the quantification of dry biomass production. For the cultivation of Chlorella sp., no major differences were observed between the two tanks. The lipid content of S. acuminatus was 17.64%, similar to Chlorella sp., but Nannochloropsis sp. presented a lower percentage (6.48%). It was not possible to quantify lipids for S. acuminatus and Nannochloropsis sp. in tank 2, since the amount of biomass obtained was not sufficient, a fact that can be justified by the action of external factors, which influenced the production of biomass and consequent lipid accumulation.

Analyzing the results, it is clear the impact that the action of factors such as temperature and light, mainly, has on the production of microalgae biomass. It is important to note that, with each cultivation carried out, the properties of the medium may vary, leading us to take into account that the characteristics of groundwater and septic tank effluents vary even within the same region and may impact production efficiency. Therefore, the choice of the species to be cultivated in large-scale production must be made based on the particularities of the cultivation medium, such as the concentration of nutrients, which depends on the characteristics of the effluent from the septic tank, in addition to the salinity of the groundwater where it is observed that each well has particular physical–chemical characteristics, which depend on the region in which it is located.

Rehman et al. (2022) comment that open and closed cultivations have pros and cons for biomass production. Open ones require lower construction and maintenance costs, and it is noted that for large-scale production there is lower energy consumption. However, the risk of contamination is high and the control of growth and culture parameters such as the effect of medium mixing, pH, temperature and nutrient concentration is difficult. Closed production systems, which are composed of photobioreactors, require less space, there is high control of cultivation parameters and a low risk of contamination. The disadvantage is that increasing production is extremely difficult, and construction and maintenance costs and energy consumption are also extremely high, for large-scale production. Therefore, the lipid content results obtained in the present study confirm that the lack of control over cultivation parameters is a preponderant factor in lipid accumulation, a factor that can be decisive in large-scale production.

To carry out the cost analysis, it was necessary to establish the annual production of biodiesel and then proceed with cost sizing. When dealing with the subject Davis Aden & Pienkos (2011) established a production of 10 million gallons and an operation of 330 days per year. When carrying out the cost analysis Oliveira et al. (2022) considered production in closed cultivation, in tubular reactors. In the present study, open cultivations in raceways were considered, subjected to natural lighting, with constant agitation and with nutrients coming from septic tanks in brackish environments and a comparison of costs was made in open and closed production modes.

The equipment required for open cultivation consists of cultivation tanks, pumps, lipid storage tanks, pre-treatment reagents, materials for lipid extraction and separation membranes. Oliveira (2021) also shows that capital expenditure concerns indirect costs, which include work and construction, development of the cultivation site, as well as annual maintenance of the area. Davis et al. (2011) also list the annual indirect labor cost as being 25% of operating costs and other indirect costs estimated at 10% of capital costs. Nagarajan et al. (2013) assume that equipment depreciation is 10% per year. Table 3 shows the capital expenditure for open crops calculated based on the work of Oliveira (2021).

It is observed that these authors carried out calculations of operating costs for closed crops, with temperature and light control. The large-scale operation costs carried out for the present study took into account open cultivation, in raceways type tanks, under constant agitation, in addition to accounting for the costs of obtaining domestic effluent and drilling artesian wells to obtain the cultivation media, through mixing. Comparing the total operating costs of the present study with those carried out by Oliveira (2021), the latter proved to be 53% more expensive than the former. Thus, it can be seen that the implementation of open microalgae cultivation stations for biodiesel production purposes becomes more economically viable than the closed one. An important factor to be observed is the use of domestic effluent from septic tanks, as well as water from brackish wells, which makes the production process cheaper.

The costs of implementing the open cultivation system were based on the work of Azeredo (2012), who calculated for 200 hectares of built area. The other items were based on the work of Oliveira (2021), who projected the costs for an annual production of 10 million gallons. The costs of drilling artesian wells and obtaining domestic effluents were based on market research carried out.

It is important to highlight that several variables must be observed to quantify the costs of biodiesel production, namely factors such as average annual inflation, price of biodiesel, price of glycerin (byproduct of the transesterification reaction), yield of the transesterification reaction and tariff annual average of industrial electrical energy. The price of biodiesel was estimated based on the average sales price in Brazil by the Agência Nacional de Petróleo (ANP) in 2020, which was R$2.459/L. Likewise, glycerin has an estimated price of R$833.00/ton (Oliveira 2021).

The cost estimates of the present work were made based on the implementation of an open cultivation system in raceway lagoons, with this mode of production being responsible for 19% of the total cost of the cultivation system. The disparity in values between open and closed crops is noticeable, with the former being much cheaper than the latter. However, factors such as productivity of biomass and lipid production and control of operating parameters that considerably affect microalgal growth must be observed when producing biomass on a large scale for bioenergy purposes.

An important aspect to be observed is that the annual biomass production volumes and built areas used in this work may change, depending on the conditions and energy demand of the region where the biodiesel production station will be implemented. There is the possibility of building these production systems close to small and medium-sized cities, with the possibility of reusing part of the domestic effluent from septic tanks produced in these locations and redirecting them for use as a source of nutrients for the production of microalgae biomass.

It is also observed that the semi-arid region, due to its scarcity of surface water and low rainfall, uses groundwater to meet part of the region's water demand. However, most of these waters have high salinity, requiring treatment through the desalination process. Thus, in the implementation of large-scale production systems, in addition to well water, waste from the desalination process can be used, which generates a water stream with a salt concentration greater than that of the supply stream, making it possible, in large-scale cultivation, to mix raw water from artesian wells with the concentrate from the desalination process.

Through cost analysis, it can be seen that the implementation of microalgae biomass production stations for bioenergy purposes still proves to be an expensive process, both with regard to the construction, implementation and maintenance of the physical structure, and the harvesting processes of the biomass, lipid extraction and transesterification reaction. Hence the importance of the present study, which sought to reduce production through the reuse of black water, coming from septic tanks, which proved to be an important source of nutrients. This study also used brackish water from wells, with the aim of enhancing lipid accumulation, increasing the productivity of biodiesel production, as well as ceramic membranes as a filtering medium for the separation of biomass and the cultivation medium. Such membranes have the advantage of low production costs, in addition to the possibility of reusing them numerous times.

From the training curves of the neural networks, it was possible to determine that there was adaptation of the networks in relation to the kinetic data. However, due to the high experimental error, there is a perception that there was overtraining of them. In this way, it is possible to infer that neural networks can represent the growth kinetics of microalgae adequately, however, in order to obtain concrete results, it is necessary to carry out controlled studies (temperature, luminosity, humidity, etc.), which in open cultivation becomes unfeasible.

Figure 2 also shows the correlation graphs between the experimental growth data and those obtained by the neural networks, with the x axis being the cultivation time in days and the y axis being the cell concentration for each day of cultivation. So much for Chlorella sp. (a) as for S. acuminatus (b), in both tanks, the correlation coefficient was 0.99, showing a high fit between the experimental data and the neural networks. As for Nannochloropsis sp. (c), in the first graph the coefficient was lower (0.96) in the first tank, due to external interference at the point corresponding to the sixth day of cultivation. This fact shows the lack of control over climatic, environmental factors and media characteristics, in relation to factors such as pH, dissolved oxygen, among others, which were evidenced in the R² value. This fact is proven in Figure 2 (C-II); when the respective point referring to day six is removed, there is an increase in the correlation coefficient.

The extreme precision obtained by the neural networks in relation to the data indicates that there is overtraining of them, since the high margin of error of the experimental data does not allow the points to be given with the high precision necessary for the correct training of the networks. In order to properly train networks with data of this nature, a much larger number of experiments would be necessary, which makes the training process unfeasible in a timely manner.

We can observe that the networks showed good learning in carrying out the lag, log and death phases, correlating the points appropriately. The points in the stationary phase showed behavior similar to a parabola, which can be observed in the graphs in Figures 2(a)–2(c), due to the high experimental error presented by the experiments as well as the low number of experimental points used. In this way, we can determine that carrying out low-control experiments was decisive for the failed training of ANNs. Even considering the flaws, the correct representation of the ANNs of the lag, log and death phases shows satisfactory behavior for research and is an indicator that, in controlled experiments and with a larger volume of data, it is possible to generate correct curves with machine learning for microalgal growth kinetics.

From the results obtained, in general, it is concluded that the production of microalgae biomass in open cultivation, using domestic wastewater as the exclusive source of nutrients, is an alternative for large-scale production, even though the efficiency of production is lower compared to closed cultivation, in which it is possible to control cultivation parameters. The closed mode of production, however, has much higher costs, which makes large-scale production unfeasible. Thus, the production cost factor appears to be preponderant for large-scale biomass production for bioenergy purposes, and the present study presents a viable alternative that can contribute to supplying part of the energy demand. Furthermore, the environmental factor can be highlighted, as the present study showed high levels of nutrient removal and a reduction in the polluting load, as well as a reduction in contamination levels in the environment, through a reduction in the number of bacteria in the environment. E. coli group is the main indicator of fecal contamination in water bodies.

The present study showed that the lack of control over cultivation parameters has a direct impact on the lipid accumulation of microalgal cells. The risk of contamination, changes in pH and temperature, as well as nutrient concentration impact in a way that reduces the lipid content, compared to closed cultivation systems. The species that presented the highest values was Chlorella sp. (18.39% and 20.52%), followed by S. acuminatus (17.64%). For this species and Nannochloropsis sp. not enough biomass was obtained for lipid quantification in the second culture tank. Therefore, for large-scale crops, it is important to search for strategies that minimize the lack of control over the parameters of the production process, which, as proven in this work, bring negative impacts to the production system.

The study showed that the open cultivation of microalgae using domestic effluents as a source of nutrients can have significant positive environmental impacts. First, this practice can help mitigate water pollution, as the nutrients present in effluents, such as nitrogen and phosphorus, are used by microalgae for their growth, thus reducing the load of pollutants in receiving bodies of water. Furthermore, the cultivation of microalgae can contribute to the capture of carbon dioxide (CO₂), helping to reduce greenhouse gas emissions. Microalgae have a high rate of photosynthesis and can absorb atmospheric CO₂ during their growth, which helps combat global warming. Additionally, the cultivation of microalgae can result in the production of biomass that can be used for the production of biofuels, reducing dependence on fossil energy sources and, therefore, contributing to the mitigation of climate change. In summary, growing microalgae with domestic wastewater as a source of nutrients has a range of environmental benefits, from reducing water pollution to carbon capture and the production of renewable biofuels.

Throughout the study it was observed that the concentration of nutrients directly affects the obtaining and efficiency of biomass production. In the case of Chlorella sp. the low initial concentrations of nutrients such as ammonia may have positively affected the growth kinetics and consequently obtained greater biomass production per liter of cultivation, since the lipid percentages and biomass production levels per liter of cultivation were the highest among the three species. The others (S. acuminatus and Nannochloropsis sp.), which were grown in media with higher concentrations of the same nutrient, achieved lower lipid percentages and biomass production. Thus, media with high concentrations of nutrients can inhibit cell growth, which is a function of the species that is being produced. Therefore, for large-scale crops, it is important to study the kinetic characteristics of each species depending on nutrient concentration, in order to maximize production.

For the three species under study, the ammonia removal percentages were quite satisfactory. For S. acuminatus and Nannochloropsis sp. such results were greater than 90% and for Chlorella sp., they were greater than 70%. Regarding the reduction in BOD₅, percentages greater than 80% were observed in S. acuminatus and greater than 50% in other species. From these results, it appears that the use of the studied means can bring positive environmental impacts, by reducing the levels of pollution in water bodies. It was observed that in media with higher concentrations of ammonia the occurrence of nitrification was favored, due to the action of nitrifying bacteria, resulting in an increase in nitrate and nitrite levels.

The preliminary study of the biomass separation efficiency using ceramic membranes showed that this process can be quite efficient in large-scale production. Such membranes have the advantages of low production cost and high durability, and in large-scale processes they can generate major impacts in reducing production costs. From visual analyses of the color of the permeate and concentrate, it was noticed that the vast majority of microalgae were retained in the concentrate stream. It is necessary to further study this technology, in order to seek optimization of the system and measurement variables, increasing biomass retention and shorter filtration time.

The production of biomass using domestic effluent proved to be efficient in decontaminating this environment. This fact was proven by the decrease in the number of colonies of the bacterium E. coli, and it was observed that in Chlorella sp. there was a maximum decrease of 3logs in one of the tanks. For S. acuminatus and Nannochloropsis sp. the observed decrease was 2 log, between the initial and final concentrations. In all cultivations, involving the three species under study, it was verified that the cultivation of microalgae from domestic effluents can also bring environmental benefits in the sense of reducing contamination rates in water bodies.

The analysis of large-scale production costs in open crops proved to be 53% cheaper than closed crops. The main factors that influenced the reduction in costs, compared to the closed system, were the use of domestic effluent as a source of nutrients and brackish well water as a cultivation medium. Added to this, the proposal to use ceramic membranes in the biomass separation process can significantly contribute to reducing costs. These results highlight that, although open cultivations obtain lower intracellular lipid levels, the costs of implementing and producing large-scale cultivation systems are a decisive factor in choosing this mode of production. Despite the above, the cost analysis also shows that large-scale biomass production is still quite expensive, with regard to the construction, implementation and maintenance of cultivation systems. Therefore, it is vitally important to research and improve alternative technologies that seek to reduce such costs.

The generation of microalgal kinetic curves with ANNs proved to be unsatisfactory in terms of results. The low number of experimental points and the high margin of error resulted in curves that do not correctly represent the microalgal kinetics as a whole. The lag, log and death phases present correct approximate predictions, however, high error values were presented in the stationary stage, resulting in low values for the R² coefficient.

To produce an accurate ANN for determining the kinetics in an open reactor, it is necessary to carry out more experiments of the same nature and produce a robust database using the system. In addition to these points, it is necessary to observe the behavior for an extended period of time, in order to detect the impact of seasonal changes in environmental conditions and to determine their impact on the process over time.

With the results obtained in these experiments, it is possible to indicate the presence of predictive behavior in ANNs for predicting microalgae concentration. These results indicate that the use of ANNs for open microalgae reactors may have predictive value for the performance of microalgal kinetics. For full validation of the process, however, more robust and long-lasting experiments are still needed, with the formation of databases based on multiple experiments in periods with different seasonal impacts. Seasonalization can also be considered for carrying out the next experiments to produce more complete experimental data.

The authors would like to thank the Fundação de Apoio à Pesquisa do Estado da Paraíba (FAPESQ) for the financial support.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.