Michael D. Murphy

Comparison of modelling techniques for milk-production forecasting

The objective of this study was to assess the suitability of 3 different modeling techniques for the prediction of total daily herd milk yield from a herd of 140 lactating pasture-based dairy cows over varying forecast horizons. A nonlinear auto-regressive model with exogenous input, a static artificial neural network, and a multiple linear regression model were developed using 3 yr of historical milk-production data. The models predicted the total daily herd milk yield over a full season using a 305-d forecast horizon and 50-, 30-, and 10-d moving piecewise horizons to test the accuracy of the models over long- and short-term periods. All 3 models predicted the daily production levels for a full lactation of 305 d with a percentage root mean square error (RMSE) of ≤ 12.03%. However, the nonlinear auto-regressive model with exogenous input was capable of increasing its prediction accuracy as the horizon was shortened from 305 to 50, 30, and 10 d [RMSE (%)=8.59, 8.1, 6.77, 5.84], whereas the static artificial neural network [RMSE (%)=12.03, 12.15, 11.74, 10.7] and the multiple linear regression model [RMSE (%)=10.62, 10.68, 10.62, 10.54] were not able to reduce their forecast error over the same horizons to the same extent. For this particular application the nonlinear auto-regressive model with exogenous input can be presented as a more accurate alternative to conventional regression modeling techniques, especially for short-term milk-yield predictions.

A mechanistic model for electricity consumption on dairy farms: Definition, validation, and demonstration

Our objective was to define and demonstrate a mechanistic model that enables dairy farmers to explore the impact of a technical or managerial innovation on electricity consumption, associated CO2 emissions, and electricity costs. We, therefore, (1) defined a model for electricity consumption on dairy farms (MECD) capable of simulating total electricity consumption along with related CO2 emissions and electricity costs on dairy farms on a monthly basis; (2) validated the MECD using empirical data of 1yr on commercial spring calving, grass-based dairy farms with 45, 88, and 195 milking cows; and (3) demonstrated the functionality of the model by applying 2 electricity tariffs to the electricity consumption data and examining the effect on total dairy farm electricity costs. The MECD was developed using a mechanistic modeling approach and required the key inputs of milk production, cow number, and details relating to the milk-cooling system, milking machine system, water-heating system, lighting systems, water pump systems, and the winter housing facilities as well as details relating to the management of the farm (e.g., season of calving). Model validation showed an overall relative prediction error (RPE) of less than 10% for total electricity consumption. More than 87% of the mean square prediction error of total electricity consumption was accounted for by random variation. The RPE values of the milk-cooling systems, water-heating systems, and milking machine systems were less than 20%. The RPE values for automatic scraper systems, lighting systems, and water pump systems varied from 18 to 113%, indicating a poor prediction for these metrics. However, automatic scrapers, lighting, and water pumps made up only 14% of total electricity consumption across all farms, reducing the overall impact of these poor predictions. Demonstration of the model showed that total farm electricity costs increased by between 29 and 38% by moving from a day and night tariff to a flat tariff.

Investment appraisal of technology innovations on dairy farm electricity consumption

The aim of this study was to conduct an investment appraisal for milk-cooling, water-heating, and milk-harvesting technologies on a range of farm sizes in 2 different electricity-pricing environments. This was achieved by using a model for electricity consumption on dairy farms. The model simulated the effect of 6 technology investment scenarios on the electricity consumption and electricity costs of the 3 largest electricity-consuming systems within the dairy farm (i.e., milk-cooling, water-heating, and milking machine systems). The technology investment scenarios were direct expansion milk-cooling, ice bank milk-cooling, milk precooling, solar water-heating, and variable speed drive vacuum pump-milking systems. A dairy farm profitability calculator was combined with the electricity consumption model to assess the effect of each investment scenario on the total discounted net income over a 10-yr period subsequent to the investment taking place. Included in the calculation were the initial investments, which were depreciated to zero over the 10-yr period. The return on additional investment for 5 investment scenarios compared with a base scenario was computed as the investment appraisal metric. The results of this study showed that the highest return on investment figures were realized by using a direct expansion milk-cooling system with precooling of milk to 15°C with water before milk entry to the storage tank, heating water with an electrical water-heating system, and using standard vacuum pump control on the milking system. Return on investment figures did not exceed the suggested hurdle rate of 10% for any of the ice bank scenarios, making the ice bank system reliant on a grant aid framework to reduce the initial capital investment and improve the return on investment. The solar water-heating and variable speed drive vacuum pump scenarios failed to produce positive return on investment figures on any of the 3 farm sizes considered on either the day and night tariff or the flat tariff, even when the technology costs were reduced by 40% in a sensitivity analysis of technology costs.

Economic optimisation for a building with an integrated micro-grid connected to the national grid

This paper proposes a novel operating cost optimisation method for a building with an integrated micro-grid (MG) connected to the National Power Grid (NPG). The MG consists of a photovoltaic system (PVS) and a lead-acid battery bank (BB). The optimisation utilised a twenty-four hour forecast of building energy consumption and the corresponding electrical prices from the NPG. A piecemeal decision algorithm (PDA) and a particle swarm optimisation (PSO) algorithm were used to generate a charge/discharge rates schedule for the BB. The building energy consumption model was developed using empirical data and employee work schedules. Electricity prices were predicted using a real time pricing (RTP) model based on data from the single electricity market operator (SEM-O)[1]. The PVS and BB were modelled based on specifications from manufacturers, and weather data from the Cork Institute of Technology (CIT). The simulation results demonstrate that the building operating costs can be reduced by up to 23 % per day for a single charge/discharge rates schedule, or by up to 30 % per day for a multiple charge/discharge rates schedule.

Machine-learning algorithms for predicting on-farm direct water and electricity consumption on pasture based dairy farms

Abstract This study analysed the performance of a range of machine learning algorithms when applied to the prediction of electricity and on-farm direct water consumption on Irish dairy farms. Electricity and water consumption data were attained through the utilisation of a remote monitoring system installed on a study sample of 58 pasture-based, commercial Irish dairy farms between 2014 and 2016. In total, 15 and 20 dairy farm variables were analysed for their predictive power of monthly electricity and water consumption, respectively. These variables were related to milk production, stock numbers, infrastructural equipment, managerial procedures and environmental conditions. A CART decision tree algorithm, a random forest ensemble algorithm, an artificial neural network and a support vector machine algorithm were used to predict both water and electricity consumption. The methodology employed backward sequential variable selection to exclude variables, which added little predictive power. It also applied hyper-parameter tuning with nested cross-validation for calculating the prediction accuracy for each model on unseen data (data not utilised for model development). Electricity consumption was predicted to within 12% (relative prediction error (RPE)) using a support vector machine, while the random forest predicted water consumption to within 38%. Overall, the developed machine-learning models improved the RPE of electricity and water consumption by 54% and 23%, respectively, when compared to results previously obtained using a multiple linear regression approach. Further analysis found that during the January, February, November and December period, the support vector machine overpredicted electricity consumption by 4% (mean percentage error (MPE)) and water consumption by 21% (MPE), on average. However, overprediction was greatly reduced during the March – October period with overprediction of electricity consumption reduced to 1% while the overprediction of water consumption reduced to 8%. This was attributed to a phase shift between farms, where some farms produce milk all year round, some dry off earlier/later than others and some farms begin milking earlier/later resulting in an increased the coefficient of variance of milk production making it more difficult to model electricity and water accurately. Concurrently, large negative correlations were calculated between the number of dairy cows and absolute prediction error for electricity and water, respectively, suggesting improvements in electricity and water prediction accuracy may be achieved with increasing dairy cow numbers. The developed machine learning models may be utilised to provide key decision support information to both dairy farmers and policy makers or as a tool for conducting macro scale environmental analysis.

Monetary savings produced by multiple microgrid controller configurations in a smart grid scenario

This paper examined the annual monetary savings (€) produced by multiple microgrid controllers in an office building connected to a smart grid. Energy consumption of each occupant of the office building was recorded in order to obtain the total hourly electricity consumption of the building. Solar photovoltaic power outputs were obtained from the Cork Institute of Technology Zero2020 microgrid test bed. Four different electricity tariffs, four dynamic feed-in tariffs (FITs) and three microgrid control methods were evaluated in order to assess the most cost effective control method. The microgrid control methods were; load priority, grid priority and optimal cost priority. The annual monetary savings relative to cost of buying electricity from the grid varied from 83% to 114% across the various electricity tariffs and microgrid controllers. The results showed that varying feed-in tariff dynamically lead to greater monetary savings and optimal cost priority control produced the highest monetary savings.

A load shifting controller for Cold Thermal Energy Storage systems

A method for controlling the generation of ice for Cold Thermal Energy Storage (CTES) is presented in this paper. A load shifting controller is developed to find the optimum trajectory for CTES in a dynamic environment. The variation in ice building efficiency due to the nonlinear relationship between changing ambient air temperature and ice charge level is accommodated by employing dynamic programming in a three dimensional space. The combined system learns from its errors and adapts during simulation. The optimum ice building trajectory is progressively updated as new information comes online. The energy saving performance of the controller is assessed using a simple optimization index.

Multiple linear regression modelling of on-farm direct water and electricity consumption on pasture based dairy farms

Abstract An analysis into the impact of milk production, stock numbers, infrastructural equipment, managerial procedures and environmental conditions on dairy farm electricity and water consumption using multiple linear regression (MLR) modelling was carried out. Electricity and water consumption data were attained through the utilisation of a remote monitoring system installed on a study sample of 58 pasture-based, Irish commercial dairy farms between 2014 and 2016. In total, 15 and 20 dairy farm variables were analysed on their ability to predict monthly electricity and water consumption, respectively. The subsets of variables that had the greatest prediction accuracy on unseen electricity and water consumption data were selected by applying a univariate variable selection technique, all subsets regression and 10-fold cross validation. Overall, electricity consumption was more accurately predicted than water consumption with relative prediction error values of 26% and 49% for electricity and water, respectively. Milk production and the total number of dairy cows had the largest impact on electricity consumption while milk production, automatic parlour washing and whether winter building troughs were reported to be leaking had the largest impact on water consumption. A standardised regression analysis found that utilising ground water for pre-cooling milk increased electricity consumption by 0.11 standard deviations, while increasing water consumption by 0.06 standard deviations when recycled in an open loop system. Milk production had a large influence on model overprediction with large negative correlations of −0.90 and −0.82 between milk production and mean percentage error for electricity and water prediction, respectively. This suggested that overprediction was inflated when milk production was low and vice versa. Governing bodies, farmers and/or policy makers may use the developed MLR models to calculate the impact of Irish dairy farming on natural resources or as decision support tools to calculate potential impacts of on-farm mitigation practises.

One-day-ahead cost optimisation for a multi-energy source building using a genetic algorithm

This paper proposes strategies for operating cost optimisation of a multi-energy source building. The optimisation is based on a day-ahead forecast of building energy usage. The building in question is powered by multiple energy sources including a wind turbine, a photovoltaic system, a lead-acid battery system, and the national power grid. The optimisation method presented in this paper is a genetic algorithm. This algorithm uses the energy demand of the building, energy supplied from the wind turbine and photovoltaic system, and real-time electricity pricing to optimise the operating timetables for the batteries. Simulation results demonstrated that daily operating costs can be reduced by up to 32 % using the genetic algorithm with a fixed charge/discharge rate, and by as much as 56 % when variable charge/discharge rates are employed, in comparison to a standard decision-based strategy.

A virtual laboratory for the simulation of sustainable energy systems in a low energy building: A case study

The aim of this paper was to develop a virtual laboratory simulation platform of the National Building Retrofit Test-bed at the Cork Institute of Technology, Ireland. The building in question is a low-energy retrofit which is provided with electricity by renewable systems including photovoltaics and wind. It can be thought of as a living laboratory, as a number of internal and external building factors are recorded at regular intervals during human occupation. The analysis carried out in this paper demonstrated that, for the period from April to September 2015, the electricity provided by the renewable systems did not consistently match the buildings electricity requirements due to differing load profiles. It was concluded that the use of load shifting techniques may help to increase the percentage of renewable energy utilisation.