 Original research
 Open Access
 Published:
Threestage dayahead scheduling strategy for regional thermostatically controlled load aggregators
Protection and Control of Modern Power Systems volume 8, Article number: 19 (2023)
Abstract
Thermostatically controlled loads (TCLs) are regarded as having potential to participate in power grid regulation. This paper proposes a scheduling strategy with threestage optimization for regional aggregators jointly participating in dayahead scheduling to support demand response. The first stage is on the profit of aggregators and peak load of the grid. The line loss and voltage deviation of regulation are considered to ensure stable operation of the power grid at the second stage, which guarantees the fairness of the regulation and the comfort of users. A single temperature adjustment strategy is used to control TCLs to maximize the response potential in the third stage. Finally, digital simulation based on the IEEE 33bus distribution network system proves that the proposed threestage scheduling strategy can keep the voltage deviation within ± 5% in different situations. In addition, the Gini coefficient of distribution increases by 20% and the predicted percentage of dissatisfied is 48% lower than those without distribution.
1 Introduction
Renewable energy (RE) will be the main energy source in the future because of its green and renewable characteristics. However, with the development of RE, its own characteristics of randomness and intermittence [1] make the scheduling of the system difficult. To increase the flexibility of the traditional scheduling model, a trend that the system considers the demand side as partially controllable is inevitable. Thermostatically controlled loads (TCLs), such as air conditioning loads, heat pumps, electric water heaters, and refrigerators, can serve as indispensable demand response (DR) [2,3,4] resources, and can be progressively used to address RE variability quickly, economically and effectively [5].
Using the adjustability of TCLs to regulate the variability of renewable energy resources has important practical significance for the power grid [6,7,8]. The main purpose of TCLs participating in power grid regulation is to solve the consumption problem. Reference [9] uses maximum power tracking control to control TCLs in real time, realizes household photovoltaic consumption, and promotes the development of household distributed energy resource technology. A realtime zeroenergy potential evaluation method with the adoption of TCTR for PVACs is proposed to balance the PV generation and ACs in [10], while through the benefitdriven approach, the optimal scheduling of distributed energy resources and flexible TCLs are realized in [11]. While the above only consider the situation in which RE generation is more than loads, RE often cannot fully meet the load demand and the participation of the power grid is also needed to ensure the balance between the supply and demand of regional power when RE is being completely used.
On the demand side, the concept of aggregator is often used to represent the aggregated loads participating in power grid regulation. TCL adjustability is often used for grid frequency regulation [12, 13], and a voltage regulation strategy is proposed in [7], one which uses the adjustability of TCLs to ensure the voltage stability of the power grid. Aggregated TCLs can also be used to realize the optimal allocation of power grid regulation tasks [14,15,16,17,18,19,20,21,22,23]. TCLs are used to participate in dayahead dispatch [21,22,23], while a dayahead scheduling strategy with TCLs is proposed to achieve peak load shedding [24]. A hierarchical control strategy for TCLs is proposed to meet the scheduling requirements, ensure user comfort and reduce communication requirements in [16]. A twostage strategy is proposed to optimally dispatch the adjustable quantity provided by TCLs in [14], with the lower stage adopting the active control strategy and the upper stage establishing the energy exchange market. A realtime local electricity market for heating, ventilation, and air conditioning loads (HVACs) considering multiple distributed energy resources (DERs) is proposed in [19]. However, the application of the market mechanism will result in some aggregators not obtaining the allocated amount when there is a small number of aggregators. This will lower the enthusiasm of aggregators to participate in regulation. Furthermore, when the loads of aggregators have a certain spatial concentration, regional regulation will have a certain impact on the safe and stable operation of the power grid.
For the fairness of scheduling, and ensuring the comfort of users and reducing the impact of regulation on the grid operation, this paper proposes a threestage scheduling strategy of regional aggregators of TCLs jointly participating in dayahead scheduling. The main contributions of this paper are as follows:
Firstly, a threestage framework is proposed for TCLs, in which the TCLs are scheduled at grid, aggregator and load levels by regulating TCLs in a distributed manner. At the first stage, an electric power transaction is built for aggregated TCLs in which the total regulation is obtained when aggregators can get the maximum interest and the peak load is at the minimum. The second stage is a distribution regulation among aggregators where the change of grid operation caused by regulations is considered, and an optimal adjustment strategy is then proposed to control the aggregated TCLs to achieve the regulation objective at the third stage. Each aggregator can get regulation and interest from power system operation. The fairness of adjustment can be ensured because of the spatial distribution of TCL aggregators in the distribution network. The threestage scheduling strategy proposed in this paper has better performance in the interests of aggregators and peak load in voltage constraint than the twostage strategy.
Secondly, a single temperature adjustment strategy is proposed to control TCLs in the threestage dayahead scheduling strategy. The temperature change of each user with heterogeneous TCL can be calculated. Because of the different initial setting temperatures, load parameters and space environments, the single temperature adjustment strategy for each TCL has advantages in the comfort level compared to group adjustment. Based on the single temperature adjustment strategy, a subsidy settlement method considering the adjustment contribution of each user is proposed in the paper.
In the following, the framework of the threestage scheduling is introduced in Sect. 2. The regulatory capacity of the TCL aggregator is described in Sect. 3, and the threestage optimization model of the strategy is explained in Sect. 4. Section 5 is the case study based on the IEEE 33bus distribution power system, and the conclusion is presented in Sect. 6.
2 Framework of threestage schedule
In a smart grid, the load data of the users can be obtained by advanced metering infrastructure (AMI) in realtime. Users can control their appliances with their phones using the Smart Home Cloud Platform (SHCP). Accordingly, it is possible for loads to participate in demand response with the grid and RE. The framework is a threestage optimization strategy proposed in this paper.

(1)
In the first stage, the distributed aggregators (DAGGs) determine the adjustable range (AR) of loads within the jurisdiction and report it to the total aggregator (TAGG) which is not real and only used to represent DAGGs to participate in regulation. TAGG participates in dayahead scheduling with grid and RE under the condition that the peak of the electricity is at a minimum, the interests of the aggregator are the largest, and the total regulations of each time are obtained.

(2)
In the second stage, because of the different spatial distributions of aggregators, TAGG obtains the regulation quantity (RQ) of each DAGG when the power grid loss is at a minimum, and the voltage deviation is in the allowable range.

(3)
In the third stage, DAGG obtains the specific temperature change (TC) of each load with the best user comfort and sends the TC of each load to the users. The users regulate their TCLs with a new temperature setpoint (NTS) through the SHCP. Finally, the adjustment results (ARS) are sent to DAGGs by AMI (Fig. 1).
3 Regulatory capacity of TCL aggregator
In many studies, the virtual energy storage [7, 14], highdimensional [8], capacitance resistance [25], thermal inertia [24] and black box models [26] have been used to represent TCLs. For a single TCL, the model is always described by an equivalent thermal parameter (ETP) model [27, 28], and the indoor temperature and the building wall temperature can be written as:
where T_{i}, T_{m} and T_{o} represent the indoor temperature, building wall temperature and outdoor temperature, respectively. R_{1} and R_{2} are the equivalent thermal resistances between indoor air and outdoor, indoor air and wall, respectively. C_{a} and C_{m} are the respective equivalent heat capacities of the indoor and the wall, and Q is the heat exchange capacity of TCL. s(t) indicates the start and stop of the TCL, and is calculated as:
where T_{set} is the set temperature of the TCL and δ represents the temperature dead zone. Ignoring the wall parameters, the expression of the indoor temperature is simplified as:
where η represents the energy efficiency ratio, and P is the electric power of the TCL. The values of P can be calculated as:
Since the TCL generally works through the compressor, when the load is turned on, the electric power is the rated electric power, and when the load is turned off, the electric power is 0.
Although the adjustable potential of a single TCL is limited, aggregating TCLs can provide great regulation potential for the power grid because of the large scale and large number of TCLs. The sum of the electric power for all TCLs at t is given as:
where P_{i} is the electric power of TCL i, and s(i) represents the on–off state of the TCL i, which is related to the setting temperature of the TCL. Therefore, the power model of the aggregator can be represented by a function with T_{set}, as:
When the maximum setting temperature change is ∆T, the power of the aggregator can be described as:
Then, the maximum adjustment capacity of the aggregator can be expressed as:
The minimum adjustment capacity should be 0. Therefore, the adjustment range of the aggregator can be represented as:
Each DAGG calculates the adjustable interval according to the user load parameters and preferences in the jurisdiction area and sends the information to TAGG to participate in the day ahead optimal scheduling.
4 Threestage optimization model of strategy
Based on the framework of the threestage schedule, a threestage optimization model is built. Figure 2 shows its overall flowchart. As can be seen, the first stage calculates the total adjustment amount by a multiobjective genetic algorithm (NSGAII) when the peak load is at a minimum and the aggregator interest is at its maximum. The second stage distributes regulation to each aggregator when the voltage constraint is satisfied and the line loss is at a minimum, while the third stage is to calculate temperature change when the predicted percentage of dissatisfied (PPD) is at a minimum. PSO is used to solve the optimization problem in the second and third stages.
4.1 The first stage
In the first stage, an electric power transaction is built, and the minimum peak load of the power grid after TCL regulation is taken as the optimization goal to ensure the safe and stable operation of the power grid, expressed as:
where P_{base, i} represents the power of other loads, and P_{ac, i} represents the power of TCLs at time i. The power constraint is shown as:
The second goal of the first stage is to obtain the greatest benefit for the TCL aggregators. According to [29], the electricity price of demand response electricity is the realtime electricity price of the demand side response market, which has a linear relationship with the total load, as:
where Ep_{dr} represents the realtime electricity price of the demand side market over a certain period of time, P_{all} represents the total load in the network at this time, while a and b represent positive electricity price coefficients. The subsidy of the aggregator to users is a quadratic function of the demand side response reduction, given as:
where B represents the total subsidy given to users by the aggregator, k is the subsidy cost coefficient of the aggregator to users, and P_{dr} represents the demand side response at this time. L_{before} and L represent the total load before and after the demand side response, respectively. L_{base} represents the basic load of the power grid, and L_{ac} represents TCLs’ power.
From (14) and (15), the optimization objective function can be described as:
where t_{1} and t_{2} are the respective start and end times of using TCLs. The inequality constraint should be considered, i.e.:
where P_{dr, t, min} and P_{dr, t, max} are the minimum and maximum adjustable capacities of the aggregator at a certain time, respectively.
Finally, because of the advantages of fast running speed and good convergence of solution sets, the multiobjective genetic algorithm NSGAII [30] is used to solve the double objective optimization model, and the adjustment amount, total benefit and total user subsidy are obtained.
4.2 The second stage
After obtaining the adjustment amount of the total aggregator, the next step is to allocate the adjustment amount to each aggregator. The TCLs that can be regulated by each aggregator have certain spatial differences, so the impact of regulation behavior on power grid operation should be fully considered during distribution. For branch connecting points m and n, the power flow is calculated as:
where v_{m} and v_{n} represent the respective squares of the voltage amplitudes of nodes m and n, r_{mn} and x_{mn} are the respective resistance and reactance of line m–n, while P_{mn} and Q_{mn} represent the active power and reactive power on node m, respectively. l_{mn} is the square of the current amplitude of line m–n.
The grid total line loss P_{loss} can be expressed as:
The optimization objective function of this stage can be described as:
The node voltage deviation constraint is shown as:
where vdiff_{i} represents the percentage voltage difference before and after adjustment of the i^{th} node, while v_{i, t} and v_{i, o} represent the voltages of node i after and before adjustment, respectively. Finally, the multidimensional particle swarm optimization (PSO) algorithm [31] is used to obtain the adjustment of each aggregator at each time because the solution variables are multidimensional.
4.3 The third stage
In terms of controlling TCLs under the jurisdiction of a single aggregator, an optimal adjustment strategy is proposed. From the user perspective, the optimization focus is on the comfort of users. The Fanger thermal comfort model is a common model used to quantify the discomfort of the user and describe the dissatisfaction of the user with room temperature. Since the discomfort is related mainly to the indoor temperature, other factors are assumed to be constants, and then the interpolation method is used to fit the discomfort level of the user and setting temperature. The fitted expression is simplified as [32]:
where the predicted percentage of dissatisfied (PPD) represents the dissatisfaction of the user with the indoor temperature, a is a positive proportional coefficient, the temperature setting value after adjustment is Tk set, and T0 set represents the expected temperature value of the user, which is the initial TCL setting value. Therefore, the discomfort level of a single user χ is expressed as:
Users have different expected temperatures, resulting in different discomfort levels at the same temperature. Therefore, the discomfort level of each user needs to be calculated separately, and the user adjustment behavior also needs to be calculated separately. The optimization goal at the user level is that the overall discomfort level of users is the lowest, which is expressed as:
The temperature constraint is shown as [28, 33]:
In the previous content, the influence of temperature regulation on the aggregate power of TCL was analyzed. Therefore, it is also necessary to consider the aggregated power constraint of TCL during optimizing after the change in the temperature setting value.
Because the TCLs under the jurisdiction of the aggregator are heterogeneous, the fitting rate between the daily load curves after the temperature setting value is changed, and the upper optimal daily load curve is taken as the power constraint of the user level optimization model. The curve fitting rate R_{new} in this paper is greater than 0.9, i.e.:
and the fitting rate calculation is given as [36]:
where y_{i, t} and y_{t} represent the actual and target values at time t, respectively.
Because of the diversity of user regulation behavior, the current set temperature regulation of TCLs is an integer, and the specific temperature regulation of each user needs to be obtained. Therefore, integer programming based on the PSO [31] algorithm is introduced to solve the userlevel optimization model. The particle position update formula of PSO is given as:
where \(g(t)\) is the direction vector of the particle.
After obtaining the temperature regulation, how to get the subsidies of users becomes a new problem. For user i under the jurisdiction of an aggregator, this paper takes the average adjustment amount of N users with the same user preference and load at each time under each temperature adjustment amount as the contribution of the user under the corresponding temperature adjustment amount [37], expressed as:
where u_{i} is the contribution of user i, T_{r} is the temperature regulation, and N is 1000 in this paper. Then, the subsidy of user i can be expressed as:
where C_{users, i} is the subsidy of user i, m is the quantity of users, and C_{users, sum, r} is the sum subsidy of DAGGs.
5 Case studies
Based on the IEEE 33bus distribution network system, RE and three aggregators are assumed to be in the areas that have different jurisdictions, as shown in Fig. 3. The prediction of RE generation and basic loads of the load regulation point excluding the TCL and the outdoor temperature change are shown in Fig. 4. The 1000 TCLs that are mostly air conditioning loads are assumed to be involved in the regulation of each DAGG, and the load parameters meet the distribution shown in Table 1.
Using Monte Carlo simulation, 1000 groups of load parameters under each aggregator are obtained. Then, TAGG uses the maximum regulation capacity of each aggregator to participate in the dayahead scheduling. The coefficients a and b in (14) are 8.25 × 10^{–5} and 0.2 in this paper, while k is 0.00045 CNY¥/kW^{2} [23].
5.1 DAGG with different control points
For the threestage scheduling strategy shown in Sect. 4, the simulation results of the threestage optimization model are shown in Table 2. First, the power from the grid and optimal daily loads can be obtained by solving the optimization model of the first stage, as shown in Fig. 5, where the “gamultiobj” function in Matlab is employed to implement NSGAII.
The peak load of the optimal daily load curve is 14406.061 kW, and the aggregators provide 1457.283 kW peak shaving capacity for the power grid, leading to a saving of 12,241 kWh electricity. The benefit obtained by the aggregators is CNY¥ 7588.988, and the user subsidy is CNY¥ 11,386.5128. At this time, the total power regulations of the aggregators are shown in Table 2.
Different locations of TCL control points can lead to different results in the second stage. While the different results indicate the small difference of adjustment values, the change trend of regulation of each DAGG is the same. In this section, case 1 and case 2 with different control points are employed to study the effectiveness of the proposed strategy. The TCL control points of each DAGG are shown in Table 2. Then, the best regulation of each DAGG is obtained in Table 2 by a 3 × 12 dimensions PSO algorithm while the outputs are lower than the regulatory capacity of each TCLs aggregators when the optimal operation of the grid is considered.
Comparing the regulations of case 1 and case 2, as shown in Fig. 6, the regulations have differences from 6:00 to 8:00 because of the volatility of TCLs when opening, but are almost the same at the other times, suggesting that the proposed threestage scheduling strategy can be used in different situations.
After obtaining the adjustment amount of each DAGG, the TCL temperature setting value is changed to take the best user comfort level as the optimization goal, and the PSO integer programming algorithm is used to optimize the solution, which has 1 × 1000 dimensions outputs between 0 and 3.
The temperature regulations are obtained by the optimization of the third stage. Taking DAGG1 in case 1 as an example, as shown in Fig. 7, each TCL has different regulations to ensure the thermal comfort level.
In cases 1 and 2, the threestage scheduling strategy can successfully complete the optimal scheduling, ensure the stable operation of the power grid and reduce the impact on user comfort. In general, the threestage scheduling strategy proposed is suitable for regional aggregators with different TCL control points.
However, from the TCL power curve in case 1, there are errors caused by the heterogeneity of TCLs. The errors can be reduced by using the battery configured by each DAGG in Fig. 8.
Because this paper studies TCL participation in dayahead scheduling and the use of a battery as the supplement of TCLs, the mainly affected battery characteristics are the rated power and capacity. Assuming that the price of battery output is 0.2 CNY¥/kWh, the interests of each DAGG are shown in Table 3. DAGGs can obviously receive more interest with battery participation, and the sum subsidy of users is now CNY¥ 10,720.9852.
5.2 Comparison with the twostage scheduling strategy without distribution function

A.
The twostage scheduling strategy with market
In the threestage scheduling strategy proposed in this paper, the second stage is to distribute the regulation to aggregators to ensure the fairness of regulation. This section compares the results of the threestage scheduling with distribution and twostage scheduling without distribution [6, 14, 24].
The Gini coefficient is an indicator that measures the inequality of income distribution and is used to measure the fairness of the distribution process of regulation in this paper. The closer its value is to 0, the more even the distribution is [34]. The Gini coefficient is calculated as:
where G is the Gini coefficient, n is the number of all members, and x_{i} is the data of member i.
Assuming that the regulated electricity price relationship of each DAGG is as shown in Fig. 9, the second stage in the scheduling strategy proposed in this paper is replaced by the market form. After optimal dispatching, the voltage deviations of each node of the IEEE 33bus distribution power system with or without a market are shown in Fig. 10. As shown, the voltage deviations of nodes 15–18 are larger than the largest deviation using the twostage strategy. When using the twostage strategy, the discomfort of the users of each DAGG is 9000, 2686 and 2331, respectively, and these levels are significantly higher than the discomfort of the users of each DAGG in the proposed strategy.
Using (33) to calculate the Gini coefficient of the two strategies, the Gini coefficients are 0.3510 and 0.5444 with and without distribution, respectively. Thus, it proves that using the threestage scheduling strategy proposed in this paper is clearly safer for the grid, more comfortable for users and fairer for DAGGs.

B.
The twostage scheduling strategy with voltage constraint
The results are compared with the twostage scheduling strategy with voltage constraint, which combines the first and second stages proposed in this paper into one stage. After solving the multiobjective optimization model of each first stage, the Pareto front of each strategy is shown in Fig. 11. It is clear that the peak load is lower and the total interest is greater with the threestage strategy proposed here.
Thus, the threestage scheduling strategy proposed can obtain the greater interest for aggregators and better user comfort level when the safe and stable operation of power grid is guaranteed.
5.3 Comparison with the scheduling strategy based on group adjustment
Single control is used to control TCLs of aggregators in this paper, and every TCL gets its own temperature change because of the different parameters. Group control has always been used to control TCLs [17, 35]. When compared with the group control where grouping is based on rating power, as shown in Table 4, the single control can get lower PPD, which can better ensure user comfort.
In the third stage, the different temperature regulations of each TCL are calculated in the final stage, and are used as a basis for subsidy allocation. However, for users with the same temperature change, the different setting temperatures and space environments should lead to different contributions in the regulation, resulting in different subsidies. In Fig. 12, taking 50 users of DAGG1 in case 1 as an example, users having the same temperature regulations can receive different subsidies because of the TCLs’ heterogeneity and settingtemperature difference.
6 Conclusions
A threestage dayahead scheduling strategy is proposed for TCLs. This scheduling strategy adds a distribution function among the regional load aggregators to ensure the stable operation of the power grid, guarantee the fairness of regulation among aggregators and improve user comfort. Furthermore, because of the heterogeneity of TCLs, the optimal fitting curve cannot be met entirely in the third stage. Therefore, the scheme in which aggregators configure a certain battery is proposed to reduce error of load scheduling and increase aggregator profit. The effectiveness of the proposed strategy is verified by simulation with the IEEE 33bus distribution network system.
The results show that the threestage scheduling strategy proposed in this paper can adapt to different situations with different TCL control points of aggregators. Compared with the twostage strategy not containing distribution function, the proposed strategy can better ensure the safe and stable operation of the power grid, and keeps the voltage deviation within ± 5%. Simultaneously, the Gini coefficient increases by 20%, and the comfort level of the users increases by 48% in the case study. Compared with the strategy based on group adjustment, the single temperature adjustment used in this paper can better insure user comfort. Given this, the subsidy settlement method considering the adjustment contribution of each user is then proposed.
Availability of data and materials
Not applicable.
References
Ding, Y., Xie, D., Hui, H., Xu, Y., et al. (2021). Gametheoretic demand side management of thermostatically controlled loads for smoothing tieline power of microgrids. IEEE Transactions on Power Systems, 36(5), 4089–4101.
Chen, Q., Wang, W., & Wang, H. (2022). Bilevel optimization model of an active distribution network based on demand response. Power System Protection and Control, 50(16), 1–13.
Yang, H., Zhang, X., Ma, Y., et al. (2021). Critical peak rebate strategy and application to demand response. Protection and Control of Modern Power Systems, 6(3), 357–370.
Ding, X., Xu, J., Sun, Y., et al. (2022). A demand side controller of electrolytic aluminum industrial microgrids considering wind power fluctuations. Protection and Control of Modern Power Systems, 4, 751–763.
Radaideh, A., AlQuraan, A., AlMasri, H., et al. (2021). Rolling horizon control architecture for distributed agents of thermostatically controlled loads enabling longterm gridlevel ancillary services. International Journal of Electrical Power & Energy Systems, 127, 106630.
Wei, C., Xu, J., Liao, S., et al. (2018). Coordination optimization of multiple thermostatically controlled load groups in distribution network with renewable energy. Applied Energy, 231, 456–467.
Zhang, M., & Bao, Y. (2021). Voltage control strategy for distribution network with thermostatically controlled loads equivalent energy storage model considering minimumonoff time. International Journal of Electrical Power & Energy Systems, 133, 107268.
Song, M., Gao, C., Shahidehpour, M., et al. (2019). Multitimescale modeling and parameter estimation of TCLs for smoothing out wind power generation variability. IEEE Transactions on Sustainable Energy, 10(1), 105–118.
Burmester, D., Rayudu, R., & Seah, W. K. G. (2018). Use of maximum power point tracking signal for instantaneous management of thermostatically controlled loads in a DC nano grid. IEEE Transactions on Smart Grid, 9(6), 6140–6148.
Jiang, T., Ju, P., Wang, C., Li, H., et al. (2021). Coordinated control of airconditioning loads for system frequency regulation. IEEE Transactions on Smart Grid, 12(1), 548–560.
Ross, S., & Mathieu, J. (2021). Strategies for networksafe load control with a thirdparty aggregator and a distribution operator. IEEE Transactions on Power Systems, 36(4), 3329–3339.
Mendieta, W., & Cañizares, C. A. (2021). Primary frequency control in isolated microgrids using thermostatically controllable loads. IEEE Transactions on Smart Grid, 12(1), 93–105.
Wang, Z., Bao, Y., & Di, H. (2020). Power system virtual inertia implemented by thermostatically controlled loads. IET Smart Grid, 3(4), 530–537. https://doi.org/10.1049/ietstg.2019.0246
Song, M., Sun, W., Wang, Y., et al. (2020). Hierarchical scheduling of aggregated TCL flexibility for transactive energy in power systems. IEEE Transactions on Smart Grid, 11(3), 2452–2463.
Wei, Q., Liao, Z., Song, R., et al. (2021). Selflearning optimal control for icestorage air conditioning systems via databased adaptive dynamic programming. IEEE Transactions on Industrial Electronics, 68(4), 3599–3608.
Xia, M., Song, Y., & Chen, Q. (2019). Hierarchical control of thermostatically controlled loads oriented smart buildings. Applied Energy, 254, 113493. https://doi.org/10.1016/j.apenergy.2019.113493
Gong, X., CastilloGuerra, E., CardenasBarrera, J. L., et al. (2021). Robust hierarchical control mechanism for aggregated thermostatically controlled loads. IEEE Transactions on Smart Grid, 12(1), 453–467.
Li, W., Gong, G., Ren, Z., et al. (2022). A method for energy consumption optimization of air conditioning systems based on load prediction and energy flexibility. Energy, 243, 123111.
Hui, H., et al. (2022). A transactive energy framework for inverterbased HVAC loads in a realtime local electricity market considering distributed energy resources. IEEE Transactions on Industrial Informatics, 18(12), 8409–8421.
Li, S., Peng, J., Zou, B., et al. (2021). Zero energy potential of photovoltaic directdriven air conditioners with considering the load flexibility of air conditioners. Applied Energy, 304, 117821.
Luo, F., Dong, Z., Meng, K., et al. (2017). An operational planning framework for largescale thermostatically controlled load dispatch. IEEE Transactions on Industrial Informatics, 13(1), 217–227.
Chen, P., Bao, Y., Zhu, X., et al. (2019). Dayahead scheduling of large numbers of thermostatically controlled loads based on equivalent energy storage model. Journal of Modern Power Systems and Clean Energy, 7(3), 579–588.
Cheng, L., Bao, Y., et al. (2020). A dayahead scheduling of largescale thermostatically controlled loads model considering secondorder equivalent thermal parameters model. IEEE Access, 8, 102321–102334.
Chen, Y., Luo, F., Dong, Z., et al. (2019). A dayahead scheduling framework for thermostatically controlled loads with thermal inertia and thermal comfort model. Journal of Modern Power Systems and Clean Energy, 7(3), 568–578.
Zhang, R., Jiang, T., Li, G., et al. (2021). Stochastic optimal energy management and pricing for load serving entity with aggregated TCLs of smart buildings: a Stackelberg game approach. IEEE Transactions on Industrial Informatics, 17(3), 1821–1830.
Kazmi, H., Suykens, J., Balint, A., et al. (2019). Multiagent reinforcement learning for modeling and control of thermostatically controlled loads. Applied Energy, 238, 1022–1035.
Wang, P., Wu, D., & Kalsi, K. (2020). Flexibility estimation and control of thermostatically controlled loads with lock time for regulation service. IEEE Transactions on Smart Grid, 11(4), 3221–3230.
Fan, D., Zhang Shu, S., Wang, Y., et al. (2022). Day ahead scheduling strategy for air conditioning load aggregators considering user regulation behavior diversity. Power System Protection Control, 50(17), 133–142.
Soliman, H. M., & LeonGarcia, A. (2014). Gametheoretic demandside management with storage devices for the future smart grid. IEEE Transactions on Smart Grid, 5(3), 1475–1485.
Zhou, H., Ding, M., Hu, B. et al. (2018). A NSGAII based peak load shifting optimization method for customers with distributed generators considering timeofuse price. In 2018 37th Chinese Control Conference (pp 7626–7631).
Sisodiya, S., Kumbhar, G. B., & Alam, M. N. (2018). A home energy management incorporating energy storage systems with utility under demand response using PSO. In: 2018 IEEMA engineer infinite conference (pp. 1–6).
Liu, S. (2019). Temperature control load control strategy and incentive mechanism for smart grid frequency regulation. Yanshan University.
Huang, Q., & Li, J. (2020). Research progress on the mechanism of human thermal sensation changes when the ambient temperature changes suddenly. Journal of Textile Research, 41(04), 188–194.
Kendall, M. G., & Stuart, A. (1958). The advanced theory of statistics. Hafner Publishing Co.
Chu, C., Jong, T., Huang, Y. (2007). Mitigating DLC constraints of airconditioning loads using a groupDLC method. In IEEE power engineering society general meeting (pp. 1–6), FL, USA.
Zhang, S. (2002). Discussion on the goodness of fit index of curve regression. Chinese Health Statistics, 01, 9–11.
Vanouni, M., & Lu, N. (2018). A reward allocation mechanism for thermostatically controlled loads participating in intrahour ancillary services. IEEE Transactions on Smart Grid, 9(5), 4209–4219.
Acknowledgements
Not applicable
Funding
This work was supported in part by the National Natural Science Foundation of China (No. 52007126 and No. U2166209).
Author information
Authors and Affiliations
Contributions
DF is responsible for the main work of writing manuscript. SZ (Correspondent Author) is responsible for the core innovation and revision of paper. HH is responsible for the data analysis of the model in the paper. LZ is responsible for the data analysis of the model in the paper. YW is responsible revising the grammar and other languages of the paper. XX is responsible for the core innovation of the paper. All authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Fan, D., Zhang, S., Huang, H. et al. Threestage dayahead scheduling strategy for regional thermostatically controlled load aggregators. Prot Control Mod Power Syst 8, 19 (2023). https://doi.org/10.1186/s4160102300295w
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s4160102300295w
Keywords
 Demand response
 Thermostatically controlled loads
 Threestages scheduling strategy
 Regional aggregators
 PPD
 Gini coefficient