Sediment microbial fuel cells (SMFCs) are a typical microbial fuel cell without membranes. They are a device developed on the basis of electrochemistry and use microbes as catalysts to convert chemical energy stored in organic matter into electrical energy. This study selected a single-chamber SMFC as a research object, using online monitoring technology to accurately measure the temperature, pH, and voltage of the microbial fuel cell during the start-up process. In the process of microbial fuel cell start-up, the relationship between temperature, pH, and voltage was analysed in detail, and the correlation between them was calculated using SPSS software. The experimental results show that, at the initial stage of SMFC, the purpose of rapid growth of power production can be achieved by a large increase in temperature, but once the temperature is reduced, the power production of SMFC will soon recover to the state before the temperature change. At the beginning of SMFC, when the temperature changes drastically, pH will change the same first, and then there will be a certain degree of rebound. In the middle stage of SMFC start-up, even if the temperature will return to normal after the change, a continuous temperature drop in a short time will lead to a continuous decrease in pH value. The RBF neural network and ELM neural network were used to perform nonlinear system regression in the later stage of SMFC start-up and using the regression network to forecast part of the data. The experimental results show that the ELM neural network is more excellent in forecasting SMFC system. This article will provide important guidance for shortening start-up time and increasing power output.
Microbial fuel cell (MFC) is being viewed as a potential bio-electrochemical device capable of producing energy in the form bioelectricity apart from wastewater treatment [
In recent years, researchers in various countries have studied MFC in terms of microorganisms, electrodes, configurations, matrices, operating conditions, and electrochemical properties and found that although microorganisms are the core of MFC, nonbiological factors are more important than biological factors in the production of electricity [
As far as we know, there is no study on the relationship between pH, power generation and temperature during the start-up of microbial fuel cells. Therefore, we will focus on the relationship between cathode pH, temperature, and voltage during the start-up of a microbial fuel cell. Due to the successful application of RBF neural network and ELM neural network in other biochemistry fields, we choose them to regress the nonlinear system in the later stage of SMFC start-up, and the regression network was used to predict the pH of that period.
The single-chamber microbial fuel cell was employed in this experimental device and a control experiment was conducted. Mixed culture (anaerobic sludge) was collected from Daming Lake (Ji’nan, China) which was at 5 cm under the water. The single-chamber SMFC consisted of an anode and cathode placed in a Plexiglas cylindrical chamber (purchased from Ji’nan lanyo Technology Ltd.) with a length of 49 cm and a diameter of 9 cm (empty bed volume of 2000 mL). The anode and cathode electrodes were made of plain carbon felts (purchased from Beijing Jinglong Special Carbon Technology Co. Ltd.) and pierced in several places, forming holes ~1 mm in diameter, so that water motion in the chamber was not blocked when the cathode was placed in the distilled water. Prior to use electrodes were soaked in distilled water and tested for conductivity. Wires were used for contact with electrodes and the contact area was sealed carefully with “epoxy” material. 600 mL of anaerobic sludge was used as the deposit in the anode area and 1400 mL of distilled water was used as the cathode buffer. The anode is buried 8m below the surface of the sediment, and the cathode is located 10 cm below the water level (to prevent evaporation of water), 30 cm from the surface of the sediment. The entire experiment was performed by connecting a 1 mm diameter titanium wire and an external 1000 Ω resistor to form a loop. A sediment fuel cell was built in the lab as can be seen in Figure
Schematic of laboratory SMFC configuration with sensors and data acquisition cards.
Use data real-time acquisition device to record the SMFC voltage every 1 second through the LabVIEW interface and filter the data. During the one-month test, the voltage of SMFC ascends until it arrives at a stable value. During the entire test period, the system was not supplemented with any buffers and other substrates. The temperature was measured using a LM35DZ temperature sensor (LM35 is a temperature sensor produced by NS company. It has high working accuracy and a wide linear working range). The output voltage is linearly proportional to the Celsius temperature, and it can provide a common room temperature accuracy of ±1/4°C without external calibration or fine tuning. The pH was measured using the Industrial Online pH and Redox sensors.
There are various abiotic factors affecting MFC power generation, including pH, temperature, dissolved oxygen, cathode liquid and so on. During the experiment, we chose aeration device to support SMFC, but we found that the effect of dissolved oxygen concentration on SMFC power generation is much less than that of temperature. The difference of power generation performance between SMFC in cathode chamber in aeration and natural reoxygenation is not obvious. The initial pH of SMFC is different for diverse cathode fluid, so it is impossible to compare the change trend of pH in the start-up process. To sum up, we chose the pH, temperature and voltage for correlation analysis, and studied the relationship between them during the start-up process. Statistical analysis software SPSS is utilized to analyse the correlation between temperature, pH and electricity generation performance during MFC start-up. This analysis uses Pearson product-moment correlation coefficient as a measure of correlation analysis. The Pearson product-moment correlation coefficient is commonly used in academic research to measure the strength of the linear correlation of two variables and the value is between -1 and 1. The formula is shown in
It can be seen from the Table
Correlation analysis between voltage, pH, and temperature.
Voltage [mV] | pH | Temperature [°C] | ||
---|---|---|---|---|
Voltage [mV] | Pearson Correlation | 1 | .347 | .797 |
Sig. (2-tailed) | .000 | .000 | ||
N | 628999 | 628999 | 628999 | |
pH | Pearson Correlation | .347 | 1 | .327 |
Sig. (2-tailed) | .000 | .000 | ||
N | 628999 | 628999 | 628999 | |
Temperature [°C] | Pearson Correlation | .797 | .327 | 1 |
Sig. (2-tailed) | .000 | .000 | ||
N | 628999 | 628999 | 628999 |
MFC is a complex nonlinear model, many variables and parameters in the model will affect the performance of the system. Zeng et al. [
Data-driven modelling method is based on process acquisition data. It is widely used in process industry modelling and optimization because it does not need to understand the process mechanism deeply and has strong generality of the algorithm. Neural networks have the ability to approximate arbitrary nonlinear mappings, parallel distributed computations, self-learning capabilities, and fault tolerance, and are therefore commonly used in the modelling and control of nonlinear systems. At present, there is no article modelling, simulation and prediction of the relationship between temperature, pH and voltage in the SMFC start-up process. In this paper, we choose Radical Basis Function (RBF) neural network and Extreme Learning Machine (ELM) neural network respectively to regression and prediction of the system.
RBF neural network has the characteristics of simple structure, simple training and fast learning convergence, and can approximate any nonlinear function, this paper selects the RBF neural network to achieve the regression of the nonlinear function of the last 10 days of the SMFC start-up phase. The radial basis neuron model is shown in Figure
Radial basis neuron model.
In this paper, Gaussian function is used as the radial basis function, so the activation function of radial basis neural network as
The mean square error (MSE) was used to evaluate the prediction accuracy of the model. The mathematical expression is as
Step one: get the centre of the Radical Base Function. Randomly selected The input samples are grouped according to the Nearest Neighbor rule. According to Euclidean distance between Adjust cluster centre. If the cluster centre no longer changes,
Step two: Solving the variance of Gauss function according to
Step three: Calculate the weights between the hidden layer and the output layer according to
ELM is an algorithm proposed by Huang et al for solving single hidden layer neural networks. Compared with the traditional neural network, especially the single hidden layer feedforward neural network (SLFN), ELM is faster than the traditional learning algorithm on the premise of guaranteeing the learning accuracy. For single hidden layer neural networks, ELM can initialize input weights and offsets randomly to obtain corresponding output weights. Structure diagram of single hidden layer feedforward neural network is given in Figure
Structure diagram of single hidden layer feedforward neural network.
The steps of the learning algorithm are as follows: Step one: Determine the number of hidden layer neurons and randomly set the weight Step two: Select activation function and calculate hidden layer output matrix Step three: Calculate output layer weights
The SMFC starts after the installation of the experimental equipment. Continuous operation of the system for 30 days when the power generation tends to stabilize which is considered to be the end of the start-up phase. From the analysis of Figure
The curve of voltage and power density at start-up phase of SMFC.
Because of the high frequency of data recording, this paper divides the data into three phases and is used to carefully compare the voltage, pH, and temperature changes. In the whole process of SMFC starting, the biological and electrochemical reaction in the single chamber SMFC changed the pH of the cathode, while the change trend of temperature and electricity generation is almost the same.
In the 1.5-2 days of first phase (Figure
The curve of voltage, pH, and temperature at first stage.
During the 4-7 days of the second stage in SMFC2 start-up (Figure
The curve of voltage, pH, and temperature at second stage.
In the third stage of the SMFC start-up (Figure
The curve of voltage, pH, and temperature at third stage.
From the data of the third stage, 5000 sets of good quality data were selected as training and prediction data, 4500 sets of data were randomly selected for training, and the remaining 500 sets of data were used for prediction. The input variables are time, voltage, and temperature, while the output variable is the pH value. For the RBF neural network, expected output and predictive output are shown in Figure
Comparison between expected output and predictive output.
Prediction error between expected output and predictive output.
From Figures
The relationship between the number of neurons in hidden layer, MSE, and computation time.
Number | MSE | Uptime [s] |
---|---|---|
100 | 7.7153e-07 | 0.128385 |
200 | 4.0310e-07 | 0.169729 |
500 | 3.5001e-07 | 0.678840 |
800 | 4.0323e-07 | 0.714896 |
1100 | 3.5874e-07 | 0.881071 |
1400 | 3.5081e-07 | 2.587683 |
1700 | 3.5600e-07 | 3.019430 |
2000 | 3.5780e-07 | 4.588646 |
Through the Table
Comparison between expected output and predictive output.
Prediction error between expected output and predictive output.
From Figures
There is a significant correlation between pH and voltage during the start-up of a single-chamber sediment microbial fuel cell. Changes in temperature at the initial start-up can greatly affect the change in electricity generation. A sudden increase in temperature or a sudden drop may cause a sudden increase in voltage or cut back. This has important guiding significance for increasing the efficiency of generating electricity in the early stage of microbial fuel cell start-up. In the first stage of the SMFC start-up, the pH of the cathode region changes in the same trend as the voltage and temperature change significantly, but after that, pH will rebound to a certain extent. From the second stage of the SMFC start-up, we can make a conclusion that a continuous temperature drop in a short time will lead to a continuous decrease in pH, even though the temperature will go back to normal. Although the temperature during the start-up of microbial fuel cells, pH, and voltage are significantly related to the 0.01 level, the voltage and temperature correlation is much greater than the other, which also proves that the temperature plays an important role in improving the initial production of microbial fuel cells. Microbial fuel cell as a bioelectrochemical system, SMFC system, has the characteristics of complexity, strong coupling, and various parameters cannot be obtained timely and accurately in the process of starting. At this time, it is very important to realize the regression and prediction of the system equation. By comparing the two kinds of neural networks, ELM neural network is more suitable for regression and forecasting of SMFC system. The work in this paper has important guiding significance for deeper understanding and control of the start-up and operation of microbial fuel cells.
The data used to support the findings of this study are available from the corresponding author upon request.
The authors declare that there are no conflicts of interest regarding the publication of this paper.
This work was financially supported by the National Natural Science Foundation of China (Grant: 61773226), the Key Research and Development Program of Shandong Province (Grant: 2018GGX103054, Grant: 2017GSF220005), the Project of Shandong Province Higher Educational Science and Technology Program (Grant: J18KA372), and Independent Innovation Projects of Colleges and Universities in Jinan City (Grant: 201401210).