A Load Scheduling For Smart Energy Management In Buildings With Renewable Power Generations

This article describes a Building Energy Management System (BEMS) that operates on an hourly basis and is designed to be implemented in a building, which consists of essential loads, facilitated with a utility grid connection and equipped with renewable power sources... - Asha Radhakrishnan, M P Selvan

Hike in the purchase of electrical appliances following a rising standard of living causes a growing demand for energy in domestic buildings. Inefficient use of these appliances causes wastage of energy. One way to tackle this is to give feedback to the consumers on their behaviour, which may lead to a reduction in wastage. Another way to reduce energy consumption is the application of demand side load management. The first method, even though makes the users to realise their unhealthy trend of energy utilisation, will not suggest any proper technique for them to follow – so that they could rectify the issue. Therefore, the best way to ensure the solution is to adopt the technique of Demand Response (DR).

The implementation of DR could be carried out through different methods – among which valley filling and peak load shaving directly affect the peak demand reduction. In order to accomplish this load shaping, the loads have to be scheduled properly by the user – so that heavily rated loads are not turned on unnecessarily during peak hours.
The loads are to be scheduled keeping in mind that the user’s comfort is not hindered. Meanwhile, those loads, which do not directly affect the basic comfort concern of the user, or the loads, which could be run at any time of the day, may be scheduled considering the energy availability and wastage constraints.

The loads are to be scheduled keeping in mind that the user’s comfort is not hindered…

This article describes a Building Energy Management System (BEMS) that operates on an hourly basis and is designed to be implemented in a building, which consists of essential loads, facilitated with a utility grid connection and equipped with renewable power sources.
Initially an Off-line Scheduling Algorithm is put forth that proposes an energy efficient load schedule for a day based on weather and price forecast, and as the next stage, an On-line Scheduling Algorithm is developed that runs in real time and takes into account the variations from the forecast & dynamically schedule the loads for each hour. It incorporates dynamic pricing technique, typically the Time of Use (ToU) method.

Off-line load scheduling

Load scheduling at the consumer end for energy management is a feasible option – once it is designed and executed with appropriate care suitable for the load environment. The present work tries to manage loads in a building that is supported by Hybrid Renewable Energy Systems (HRES) consisting of solar panels, wind turbines and battery along with an uninterrupted grid connection.

The renewable power generators have been designed to have an installed capacity of 20% of the total connected load in the building. A set of 4 solar PhotoVoltaic (PV) modules each of power rating 150 Wp, 5 wind turbines each of 500 W rated power output and a battery bank consisting of 4 units of 100 Ah each have been considered. With the renewable energy forecast available in prior, the off-line scheduling algorithm has been developed in steps starting with categorisation of the loads, load prioritisation and then incorporation of tariff plans at a later stage.

Load categorisation

Home appliances are initially classified into three categories, namely, appliances with real-time energy consumption mode, appliances with periodic nonreal-time energy consumption mode, and appliances with nonperiodic nonreal-time energy consumption mode.

The energy consumption of the first category of appliances is directly related to consumer behaviour, which means that after the consumer turns them ON, the appliance must be energised until they are shut down. The energy consumption of this type of appliances cannot be scheduled – and they must run immediately to satisfy the consumer’s requirements. Lights, fans, desktop computers and television are examples.

The energy consumption of the second category of appliances is periodical and fluctuant when they are in use. Air conditioners and refrigerators are examples. They could be scheduled based on maintaining the set value of temperature i.e., they need to be energised only when the temperature violates the upper or lower limits. Battery embedded devices such as laptops shall also be considered under this category.

The third type of appliances consumes energy non-periodically and does not have any specific time to run. However, they must serve their course before certain deadline. Plug-in hybrid electric vehicles and pool pumps come under this category.

The first category of devices are energised purely based on consumer behaviour – whereas the other two categories are schedulable loads.

Load prioritisation

The appliance commitment can be based upon the necessity of energy consumption by that appliance at any point of time. Again, the need or urgency of an appliance over the others should be considered for an even and efficient scheduling.

This can be ensured by allocating priority to the devices dynamically considering their status. This is applicable only to the second and third category appliances as the first category cannot be scheduled and must run immediately.

Fig. 1: Priority allocation algorithm…

Fig. 1 is the flowchart depicting implementation of the subroutine for priority allocation for the different categories of appliances.

Off-line load scheduling algorithm

The building under consideration is equipped with essential loads, a utility grid connection, and an HRES. The scheduling algorithm has been developed with the aim of maximum utilisation of the available HRES thus minimising the energy consumption from the utility grid – and hence minimising the total cost of consumption.

The basic idea of the algorithm is as mentioned hereafter. If the appliance has high priority, it should run immediately in order to satisfy the consumer’s comfort. All appliances with middle and low priorities are rescheduled based on the renewable sources’ output prediction or the electricity market price forecasting. Algorithm has been developed for any electricity tariff plans, fixed or variable. Cases with and without the incorporation of storage battery are also taken into account. This section discusses the flow of logic pertaining to a basic flat rate tariff as well as a variable tariff plan.

Fig. 2: Subroutine for load scheduling with flat rate tariff…

Fig. 2 demonstrates the implementation of a function for load scheduling without a battery in the system for a flat rate tariff. Flat rate plan employs the same cost of energy irrespective of the time or degree of consumption. Since there is no economic benefit in shifting the loads to different time slots, the focus is made mainly on the utilisation of the available renewable power. High priority appliances are run irrespective of the energy availability while middle and low priority devices are scheduled according to it.

Variable tariff plans charge differently based on the time of utilisation. This definitely encourages the consumer to schedule heavy loads during off-peak hour slots with the aim of economic savings. The algorithm developed in this work has been designed to schedule the loads according to any ToU tariff plans. This algorithm takes into account cost rate plans with any number of cost slabs and then schedules the loads accordingly. The initial step taken while dealing with multiple cost slabs is to count the number of cost slabs in the plan and to form an array of the costs in descending order. The total number of these different costs, which is the same as the size of the formed array is denoted as ‘Ncs’. An index variable ‘i’ is used to refer to each of the array element, which will also indicate whether the hour of schedule is a high cost hour or not. If ‘i’ has the value 1, then the hour is a high cost hour, if ‘i’ has a value 2, then the hour has the next highest cost, and so on. If ‘i’ has the value ‘Ncs’, it is the least cost hour. As the no. of cost slabs increases, the advantage is that the cost will be minimal for the least cost hours and maximum utilisation of these hrs can help in effective scheduling to have a minimum cost/day.

Fig. 3: Subroutine for load scheduling with ToU variable tariff plan…

The scheduling method in case of ToU cost rate plan is elaborated in Fig. 3.
The cases discussed yet do not take into account the presence of storage device in the system. As the renewable power is highly fluctuating and since there are times at which there is excess of energy availability, battery storage is inevitable.

The prime matter to be taken into account while dealing with battery is that whether it should supply the loads or it should rather be charged. In some situations, it might be ideal for the battery to stay floating. After deciding upon the functionality of battery, then the scheduling algorithm can make necessary additions and deletions to get the desired outcome.

Fig. 4: Subroutine for decision making on battery functionality with flat rate tariff…

Fig. 5: Subroutine for decision making on battery functionality with ToU variable tariff plan…

Fig. 4 and 5 illustrate the sub-functions to carry out the decision making process on fixing the battery functionality. With the inclusion of storage battery, the load scheduling begins with the decision making on the battery functionality. Later, the loads are scheduled based on the battery status and renewable power availability for each cost rate plan.

Fig. 6: Subroutine for load scheduling with battery for flat rate tariff plan…

Fig. 7: Subroutine for load scheduling with battery for ToU variable tariff plan…

Fig. 6 and 7 depict the subroutines for load scheduling including the storage battery for flat rate plan and variable plan respectively.

Overall program flow

With the sub-functions developed for the numerous possible cases, the complete off-line scheduling algorithm has been developed as illustrated in Fig. 8. Initially the tariff plan has to be assigned and the availability of battery storage in the building has to be checked. Then the load scheduling process begins with load categorisation. Once this is done, the iteration process begins.

Fig. 8: Off-line load scheduling algorithm… 

The variable ‘Tschedule’ denotes the iteration count. If the scheduling commences for 6 a.m., Tschedule shall be assigned the value 1. After generating load schedule for that hour, iteration count gets incremented and the scheduling for the subsequent hour begins. For each hour, the loads are prioritised as explained in Fig.1. Subroutine 1 executes the process. Renewable power availability for the considered hour is estimated. Next, the load scheduling is carried out for the selected case by executing the subroutines corresponding to it.

The loop is iterated 24 times corresponding to each hour of the day. When the variable Tschedule exceeds 24, the loop is terminated for the day. The scheduling is carried out assuming that the grid power is available throughout the day. The algorithm finally generates load schedule for each hour – after taking into consideration the conditions pertaining to that hour.

On-line load scheduling

The basic platform on which the on-line algorithm is developed is the same as that of the off-line algorithm – where the appliances are categorised initially. Then they are scheduled according to the availability of energy from grid, user-defined priority, renewable sources and battery – with the aim of minimising the cost.

Features to bring about dynamic changes in different parameters have been considered and augmented here. The case of unavailability of grid power for a few hours through the schedule is aforethought and decision making is carried out. Similarly, cases of dynamic changes in electricity prices and user’s choice in appliance priority are also made possible in this algorithm. Actual renewable energy availability for each hour is calculated at the beginning of the hour using real time solar irradiation and wind velocity data.

Procurement of essential real-time data

For developing dynamic load scheduling algorithm, various real time data are required, which include actual weather data, grid power status, actual temperature status of thermostatically controlled appliances and battery charge status of battery equipped devices.

The procurement of these data could be done using sensors and it could be entered and saved within a Microsoft Excel file. From these files, the data have been taken into the coded algorithm for processing.

Scheduling in the absence of grid power

The on-line algorithm is so designed to initially categorise and prioritise and check for the availability of power from the utility grid. In case of absence of grid power, it tries to schedule only the high priority loads or the critical load whichever is possible with the available renewable energy and battery energy. Only in the worst case, all the appliances are shut down. The logic is depicted in Fig. 9.

Fig. 9: Subroutine for load scheduling in the absence of grid power…

Overall program flow

The on-line scheduling algorithm begins with categorization of loads. The renewable power availability estimated, assigned tariff plan for the day and the priority of devices suggested by the off-line algorithm are loaded into the algorithm. Dynamic changes in these aspects have been considered and updated. The availability of grid power for the hour is checked, and scheduling is carried out according to the procedure depicted in Fig. 10.

Fig. 10: On-line load scheduling algorithm…

Graphical user interface

After having developed the algorithm for on-line scheduling, it has been integrated into a system that could be installed in any building by developing a Graphical User Interface (GUI), which could effectively communicate with the user.

The GUI has been designed and developed in MATLAB GUIDE. The GUI basically reads the user preferences and different settings, and displays the schedule for the coming hour.

Fig. 11: Screenshot of BEMS main window…

The opening window of the BEMS is demonstrated in Fig. 11. The user can input the load commitment of the first category devices. The grid power availability is displayed according to the status of the same for the hour. Provisions to select any tariff structure, preference of priority, and to set renewable power generator and battery storage status are also facilitated here.

The control of the BEMS can be done from this window. The RUN button starts the system and this could be run for the whole day. The DISPLAY SCHEDULE button displays the appliance schedule for the given hour.

Fig. 12: Screenshot of tariff plans window…

On activating the TARIFF STRUCTURE button on the main window – the window shown in Fig. 12 pops up. Here, the user can select the electricity tariff plan that is to be followed, from the drop down menu. A flat rate tariff and two tariff plans are already inbuilt. A provision for customised plan is also provided in addition.

Fig. 13: Screenshot of Appliance Priority Window…

The APPLIANCE PRIORITY window shown in Fig. 13 enables the user to change the priority of the devices apart from what is being suggested by the algorithm. This is facilitated only for category 2 and 3 devices – as category 1 devices are always of high priority. The drop-down boxes are where the user may edit the priorities – whereas the adjacent boxes show the priorities suggested by the algorithm.

Fig. 14: Screenshot of renewable energy converters window…

Fig. 14 shows the renewable energy converter installation in the building. The design values are displayed in the window. Provision for changing the installation details in case of a hardware change is also provided. The status of the generators could also be input here.
In case of maintenance work or poor weather conditions, if any of these need to be shut down, those changes may be reflected here. The values entered here are used for the calculation of the renewable energy availability for each hour.

Fig. 15: Screenshot of battery storage window…

The window shown in Fig. 15 displays by default the designed specifications of the storage battery equipment. Choice of connection and disconnection of the same could also be done here.

Any change in the installation could also be mentioned here, as these values are needed for the determination of the battery energy availability at any time.

Fig. 16: Screenshot of the load schedule window…

The load schedule window in Fig. 16 shows the schedule of the appliances, which is generated by the on-line scheduling algorithm. Also, it displays the grid power requirement and the cost of energy to be paid to the utility grid for the hour. It shows the cost of energy incurred for the day till that hour from the commencement of the scheduling.

The developed algorithms were coded and simulated in MATLAB. Care was taken to include appliances belonging to all three categories – so that the effect of application of the algorithm becomes more distinct. They include both schedulable and non-schedulable loads. Table 1 gives the details of the connected loads in the building along with their time of utilisation, which has been used to form the load curve with manual operation.

Off-line load scheduling algorithm

To prove the effectiveness of the developed off-line algorithm, it has been tested for various cases. Initially, the skeletal system has been considered and analysed. It has constituted just the loads in the building with utility grid as the sole source of power with flat rate tariff. The consumer behaviour has been considered for appliance commitment.

The timings for which different loads have been committed based on which the load curve and cost of energy were obtained (Table 1). With the given load pattern, the total daily cost of energy to be paid to the utility grid was estimated to be Rs. 290.16 under a flat rate tariff of Rs. 3 /unit in the absence of local renewable power generation. Considering the installed HRES, the net renewable power availability estimated is in Fig. 17.

Fig. 17: Estimated renewable power availability…

After integrating the renewable sources, different cost rate plans have been introduced in the system, and the algorithm was evaluated for each case. Results adhering to a flat rate plan and a variable plan listed in Table 2 are discussed here.

The performance of the algorithm has been evaluated in terms of load shifting, amount of grid power drawn and cost of electricity to be paid to utility per day with the two tariffs for different cases.

Fig. 18: Load curve for flat rate tariff…

Fig. 19: Load curve for ToU variable tariff plan…

Fig. 18 and 19 demonstrate how load is shifted from peak hours to off-peak hours for different cases with and without algorithm under two different tariffs.

Table 3 compares and summarises the cost to be levied in different cases – and it could be inferred that the application of the developed off-line scheduling algorithm effectively helped in minimising the cost of energy consumption with efficient utilisation of renewable power sources, which is enhanced by the presence of storage battery.

On-line load scheduling algorithm

Fig. 20: Variation in solar irradiation between forecasted and actual values…

Fig. 21: Variation in velocity between forecasted and actual values…

On-line algorithm imbibes the dynamic variations in the different deterministic parameters and reschedules the loads accordingly. Fig. 20 and 21 demonstrate the deviation of actual values of solar irradiation and wind velocity respectively from their forecasted values considered during the time of program run. The on-line schedule has been generated based on these actual values and a corresponding variation was obtained from that during the off-line schedule. Since the concept of dynamic pricing has been brought in, possibility of dynamic price changes is also considered in the frame.

Fig. 22: Variation in electricity price between forecasted and actual values for ToU variable tariff plan…

Fig. 22 shows the forecasted and actual values in electricity price for ToU tariff plan. Results obtained for the system considered in off-line scheduling algorithm after incorporating the on-line algorithm is summarised in Table 4. They justify the application of the scheduling algorithm over manual operation.

Demand response is an efficient tool in bringing about energy conservation, reduction in wastage of energy and proficient utilisation of the utility grid without stressing the existing transmission and distribution network. An attempt to implement DR through scheduling loads at the consumer end has been made in this work. This has been achieved by designing and developing a Building Energy Management System applicable for appliance commitment in a building that is facilitated with renewable power generators as well as utility grid connection. It also considered the availability of dynamic pricing scheme by incorporating ToU variable tariff plan for electricity cost. An Off-line scheduling algorithm has been developed initially, which took in the weather and electricity price forecasts, and generated an off-line schedule for the loads each day. Keeping the off-line schedule as reference, an on-line scheduling algorithm has been developed that monitors the variations in actual weather conditions and electricity price with respect to the forecasted values and cases such as change in user priority and grid power absence.

Dynamic scheduling has been brought to its integrity by designing and incorporating a Graphical User Interface with the algorithm. This has been programmed to update every hour thus scheduling the loads for that hour. The algorithm has been tested for various cases with and without the inclusion of battery storage equipment. The BEMS at the consumer end successfully resulted in the saving of energy or money or both wherever possible. It also promotes the use of renewable sources of energy that exemplifies the decentralised nature of the current electrical network.