https://www.mdu.se/

The aim of this work is to apply and evaluate different chemometric approaches employing several machine learning techniques in order to characterize the moisture content in biomass from data obtained by Near Infrared (NIR) spectroscopy. The approaches include three main parts: a) data pre-processing, b) wavelength selection and c) development of a regression model enabling moisture content measurement. Standard Normal Variate (SNV), Multiplicative Scatter Correction and Savitzky-Golay first (SGi) and second (SG2) derivatives and its combinations were applied for data pre-processing. Genetic algorithm (GA) and iterative PLS (iPLS) were used for wavelength selection. Artificial Neural Network (ANN), Gaussian Process Regression (GPR), Support Vector Regression (SVR) and traditional Partial Least Squares (PLS) regression, were employed as machine learning regression methods. Results shows that SNV combined with SG1 first derivative performs the best in data pre-processing. The GA is the most effective methods for variable selection and GPR achieved a high accuracy in regression modeling while having low demands on computation time. Overall, the machine learning techniques demonstrate a great potential to be used in future NIR spectroscopy applications.

Building energy performance modeling is essential for energy planning, management, and efficiency. This paper presents a space heating model suitable for auto-generating baseline models of existing multifamily buildings. Required data and parameter input are kept within such a level of detail that baseline models can be auto-generated from, and calibrated by, publicly accessible data sources. The proposed modeling framework consists of a thermal network, a typical hydronic radiator heating system, a simulation procedure, and data handling procedures. The thermal network is a lumped and simplified version of the ISO 52016-1:2017 standard. The data handling consists of procedures to acquire and make use of satellite-based solar radiation data, meteorological reanalysis data (air temperature, ground temperature, wind, albedo, and thermal radiation), and pre-processing procedures of boundary conditions to account for impact from shading objects, window blinds, wind- and stack-driven air leakage, and variable exterior surface heat transfer coefficients. The proposed model was compared with simulations conducted with the detailed building energy simulation software IDA ICE. The results show that the proposed model is able to accurately reproduce hourly energy use for space heating, indoor temperature, and operative temperature patterns obtained from the IDA ICE simulations. Thus, the proposed model can be expected to be able to model space heating, provided by hydronic heating systems, of existing buildings to a similar degree of confidence as established simulation software. Compared to IDA ICE, the developed model required one-thousandth of computation time for a full-year simulation of building model consisting of a single thermal zone. The fast computation time enables the use of the developed model for computation time sensitive applications, such as Monte-Carlo-based calibration methods. 

Phosphorus (P) is one of the essential nutrients for production of food. In modern agriculture, a large part of P comes from finite sources. There are several suggested processes for reuse of P from wastewater. In this paper, the energy use of direct reuse of sludge in agriculture is compared to the energy demand connected to use of mineral P and to reuse of P after thermal processing of sludge. The study is based on literature data from life cycle analysis (LCA). In the case of direct sludge reuse the sludge stabilization processes applied and the system boundaries of the LCA has a large impact on the calculated energy demand. The results though indicate that direct reuse of sludge in agriculture is the reuse scenario that potentially has the lowest energy demand (3-71 kWh/kg P), compared to incineration and extraction of P from sludge ashes (45-70 kWh/kg P) or pyrolysis of sludge (46-235 kWh/kg P). The competitiveness compared to mineral P (-4-22 kWh/kg P) depends on the mineral P source and production. For thermal processing, the energy demand derives mainly from energy needed to dry sludge and supplement fuel used during sludge incineration together with chemicals required to extract P. Local conditions, such as available waste heat for drying, can make one of these scenarios preferable.

To be able to fulfill the Paris agreement regarding anthropogenic greenhouse gases, all potential 12 emissions must be mitigated. Wastewater treatment plants should aim to eliminate emissions of the 13 most potent greenhouse gas, nitrous oxide. In this study, these emissions were measured at a full-scale 14 reject water treatment tank during two different operation modes: nitrification/denitrification (N/DN) 15 operating as a sequencing batch reactor (SBR), and deammonification (nitritation/anammox) as a moving 16 bed biofilm reactor (MBBR). Nitrous oxide was measured both in the water phase and in the off-gas. The 17 treatment process emitted significantly less nitrous oxide in deammonification mode 0.14-0.7 %, 18 compared to 10 % of Total Nitrogen in N/DN mode. The decrease can be linked to the change feeding 19 strategy, concentration in nitrite, load of ammonia oxidized, shorter aeration time, no ethanol dosage 20 and the introduction of biofilm. Further, evaluation was done how the operational pH set point 21 influenced the emissions in deammonification mode. Lower concentrations of nitrous oxide was 22 measured in water phase at higher pH (7.5-7.6) than at lower pH (6.6-7.1). This is believed to be mainly 23 because of the lower aeration ratio and increased complete denitrification at the higher pH set point.

The aim of this paper is to evaluate and compare the performance of two machine learning methods, Gaussian process regression (GPR) and Gaussian mixture models (GMMs), as two possible methods for monitoring the sludge profile in a secondary settler tank (SST). In GPR, the prediction of the response variable is given as a Gaussian probability density function, whereas in the GMM the probability density function is built as a weighted sum of Gaussian distributions. In both approaches, a residual is calculated and a fault detection criterion is implemented via a recursive decision rule. As case study, GMM and GPR were tested using real data from a sensor measuring the suspended solids concentration as a function of the SST level in a wastewater treatment plant in Bromma, Sweden. Results suggest that GMM gives a faster response but is also more sensitive than GPR to changes during normal conditions.

In this study a natural mix of microalgae grown in wastewater of municipal character was co-digested with sewage sludge in mesophilic conditions, in both batch and semi-continuous modes. The semicontinuous experiment was divided into two periods with OLR 1 (Organic Loading Rate) of 2.4 kg VS m³ d^-1 and HRT1 (Hydraulic Retention Time) of 15 days, and OLR2 of 3.5 kg VS m³ d^-1 and HRT2 of 10 days respectively. Results showed stable conditions during both periods. The methane yield was reduced when adding microalgae (from 200 ± 25 NmL CH₄ g VSin^-1 , to 168±22 NmL CH₄ g VSin^-1). VS reduction was also decreased by 51%. This low digestability was confirmed in the anaerobic batch test. However, adding microalgae improved the dewaterability of the digested sludge. The high heavy metals content in the microalgae resulted in a high heavy metals content in the digestate, making it more difficult to reuse the digestate as fertilizer on arable land. The heavy metals are thought to originate from the flue gas used as a CO₂ source during the microalgae cultivation. Therefore the implementation of CO₂ mitigation via algal cultivation requires careful consideration regarding thesource of the CO₂-rich gas.

A photobioreactor (PBR) contains microalgae which under illumination consume carbon dioxide and substrate dissolved in water, and produce oxygen. The process is used in water recovery resource facilities with a continuous flow of wastewaster through the PBR. With several PBRs in series the reduction of substrate can be improved. This paper contains a thorough analysis of a model of PBRs in series, where each PBR is modelled with a system of three ordinary differential equations for the concentrations of dissolved substrate and biomass (algae), and the internal cell quota of substrate to biomass. Each PBR has a certain volume and irradiation. The absorption rate of substrate into the cells is modelled with Monod kinetics, whereas the biomass growth rate is modelled with Droop kinetics, in which both a minimum and a maximum internal cell quota are assumed. The main result is that the model has a unique stable steady-state solution with algae in all PBRs. Another stable steady-state solution is the wash-out solution with no algae in the system. Other steady-state solutions are combinations of these two with no algae in some of the first PBRs and algae in the rest of the PBRs in the series. Conditions on the illumination, volumetric flow and volumes of the PBRs are given for the respective solution. Numerical solutions illustrate the theoretical results and indicate further properties.

The hindered-settling velocity function for the modelling, simulation and control of secondary settling tanks can be determined from batch tests. The conventional method is to measure the velocity of the descending sludge-supernatant interface (sludge blanket) as the change in height over time in a vessel with constant cross-sectional area. Each such experiment provides one point on the flux curve since, under idealizing assumptions (monodisperse suspension, no wall-effects), the concentration of sludge remains constant just below the sludge blanket until some wave from the bottom reaches it. A newly developed method of estimation, based on the theory of nonlinear hyperbolic partial differential equations, is applied to both synthetic and experimental data. The method demonstrates that a substantial portion of the flux function may be estimated from a single batch test in a conical vessel. The new method takes into consideration that during an ideal settling experiment in a cone, the concentration just below the sludge blanket increases with time since the mass of suspended solids occupy a reduced volume over time.

Biofilm fouling is known to impact the data quality of sensors, but little is known about the exact effects. We studied the effects of artificial and real biofilm fouling on dissolved oxygen (DO) sensors in full-scale water resource recovery facilities, and how this can automatically be detected. Biofilm fouling resulted in different drift direction and bias magnitudes for optical (OPT) and electrochemical (MEC) DO sensors. The OPT-sensor was more affected by biofilm fouling compared to the MEC sensor, especially during summer conditions. A bias of 1 mg/L was detected by analysing the impulse response (IR) of the automatic air cleaning system in the DO sensor. The IR is an effect of a temporal increase in DO concentration during the automatic air cleaning. The IRs received distinct pattern changes that were matched with faults including: biofilm fouling, disturbances in the air supply to the cleaning system, and damaged sensor membrane, which can be used for fault diagnosis. The results highlight the importance of a condition based sensor maintenance schedule in contrast to fixed cleaning intervals. Further, the results stress the importance of understanding and detecting bias due to biofilm fouling, in order to maintain a robust and resource efficient process control.