CN112838580A - An Improved Bilinear Aggregation Model for Heterogeneous Temperature Control Loads and Its Distributed Hierarchical Multi-objective Coordinated Control Method - Google Patents

Abstract

The invention provides an improved heterogeneous temperature control load bilinear polymerization model and a distributed layered multi-target coordination control method thereof. Firstly, improving an original bilinear polymerization model by using a temperature control load second-order thermodynamic model, optimizing the problem of accumulated errors existing in the bilinear model, and deriving a higher-precision improved heterogeneous temperature control load bilinear polymerization model; secondly, a distributed layered multi-target control strategy based on the combination of consistency control and reverse control is provided, a plurality of temperature control load clusters in the system are regulated and controlled aiming at system power fluctuation exceeding AGC economic dispatching, and stability control services such as system load tracking and power angle and frequency are provided. The invention has the novelty that an original temperature control load bilinear polymerization model is improved, the model description precision is improved, and a distributed layered multi-target control strategy with good performance is designed, so that the TCL can better participate in the regulation and control of the power system.

Description

An improved bilinear aggregation model for heterogeneous temperature control loads and its distributed hierarchical multi-objective coordinated control method

technical field

The invention belongs to the field of power system auxiliary services and demand side response, and relates to an improved modeling and scheduling control strategy for temperature-controlled loads, in particular to an improved bilinear aggregation model of heterogeneous temperature-controlled loads and its distributed hierarchical multi-objective coordination Control Method.

Background technique

The stability and control of power angle and frequency are the guarantee for the safe and stable operation of the power system, and it is closely related to the balance of generated power in the system. Traditionally, the power system adopts the method of power generation to track the load to meet the power balance and stability of the system, and the load is regarded as a passive physical terminal. When it is still difficult to maintain the stability of the system or it is expensive to allocate the output of the generator set in the traditional way, the current load shedding/power supply measures will have a large negative impact on society. And as the power load continues to climb, a large number of intermittent power sources are centrally connected to the power system, and the proportion of large-capacity supercritical units in the system continues to increase, the ability of the power generation side to flexibly adjust the output gradually weakens. The demand for source-grid-load interactive operation makes the use of existing resources on the demand side to supplement traditional power generation dispatching to carry out auxiliary services such as power frequency regulation and peak regulation, which has attracted widespread attention.

Thermostatically controlled load (TCL), represented by air conditioners, refrigerators, and water heaters, has become one of the main research objects of flexible loads due to its advantages of rapid response, energy storage, and high controllability. When the grid load fluctuates greatly and the system reserve capacity is insufficient, TCL with abundant reserves can be used to supplement the system AGC power regulation capability, quickly maintain the system balance, and improve the safety and economy of grid operation. However, due to the characteristics of small capacity, large quantity, scattered distribution and strong response randomness of TCL monomers, it is difficult for the dispatch center to obtain information on its aggregated power consumption and response potential. Therefore, how to effectively utilize these resources is the main challenge for the dispatch center at present. challenge.

Load system modeling is the basis for TCL to participate in demand response. The aggregation model can summarize the operating characteristics of large-scale load groups and guide relevant agencies to formulate control strategies. TCL modeling has always been one of the research hotspots of scholars at home and abroad. Among them, a control-oriented TCL bilinear aggregation model can effectively reduce the calculation amount of the TCL model, avoid the dilemma of dimensionality disaster, and provide a large-scale TCL aggregation scheduling. An efficient approach used by many research applications. However, the factors considered by the current original bilinear model are too single, and there are accumulated errors, and the description accuracy of the model needs to be further improved. In addition, in order to reduce the amount of calculation in the process of establishing the aggregation model, many scholars often assume that all load system model parameters are completely consistent, which not only violates the actual situation, but also the lack of load parameter diversity often causes oscillation problems, which is not conducive to the effective control of TCL. The aggregated model of parameter diversity often produces a natural damping, which can make the system have better stability in the process of regulation. But this will inevitably lead to an increase in the number of aggregates, which leads to another problem: how to coordinate the regulation of many aggregates. According to the decision position of the control signal, the control modes of the TCL demand response signal can be mainly divided into two categories: one is centralized control, that is, the unified control of the TCL, which is characterized by high reliability and strong predictability, but because the control center It is necessary to set up communication lines with all load clusters, and there are problems of high investment cost and communication delay. The second is distributed control, which can respond quickly according to the monitoring results and its own conditions, so its response speed is fast and has high sensitivity. The problem of under-response or over-response.

SUMMARY OF THE INVENTION

In view of the above problems, the present invention aims to overcome the shortcomings of existing research, and makes the following two innovations: one is to optimize the original TCL bilinear aggregation model, and propose an improved heterogeneous TCL bilinear aggregation model. The second is to design a distributed hierarchical multi-objective coordinated control strategy for coordinating multiple TCL aggregates based on the principles of consistency control and reverse thrust control, and realize the power system power angle under the constructed inertial center system. , multi-index regulation of frequency and power.

In order to achieve the above object, the technical scheme that the present invention takes is as follows:

及初始物质温度T_m0下的平均负荷传递率

得到异质温控负荷双线性聚合模型。First, a second-order differential equation equivalent thermodynamic parameter model (Equivalent Thermal Parameter, ETP) is used to describe a single TCL, and the indoor temperature and indoor material temperature change are the two state variables observed in the ETP model, so it is also referred to as different The mass model mainly considers the heat exchange process between indoor materials and indoor air, and approximates the load transfer rate α _on/off of each temperature interval to the desired temperature set value

and the average load transfer rate at the initial material temperature T _m0

A bilinear aggregation model with heterogeneous temperature-controlled load was obtained.

中的

用实时物质温度T_m(t)替换

中的T_m0，最终推导得到所述改进异质温控负荷双线性聚合模型。所提出的改进模型综合体现了温度设定值变化对温控负荷的实时影响、累积影响，以及异质性对温控负荷的影响。Then optimize the cumulative error caused by the reduction of α _on/off during the derivation of the bilinear aggregation model, and replace it with the real-time temperature set value T _set (t)

middle

Replace with real-time material temperature T _m (t)

T _m0 in the final derivation to obtain the improved heterogeneous temperature-controlled load bilinear aggregation model. The proposed improved model comprehensively reflects the real-time and cumulative effects of temperature setpoint changes on the temperature-controlled load, as well as the effect of heterogeneity on the temperature-controlled load.

Secondly, a distributed hierarchical multi-objective coordinated control method is designed for the aggregate control of multiple temperature-controlled loads.

Among them, the consistency control is applied to the aggregation quotient level to balance the output of each temperature-controlled load aggregate. The control signal λ ₀ (t) of the control center is constructed based on the PI control idea, and the output power coefficient of each temperature-controlled load is λ _i ( t) Iterate according to the consistency control. As long as the system power deficit ΔP≠0, λ ₀ (t) and λ _i (t) will iterate until the system power is balanced, and the value of each λ _i (t) is in the iterative calculation process. It can be ensured that each aggregate is assigned an output power tracking value corresponding to its own maximum power ratio.

The reverse thrust control is applied to the temperature-controlled load aggregate level, which is used to control the output power of the temperature-controlled load to track the output quota allocated by the load aggregater; The control principle, given the reference value of the power angle δ _ref , deduce the angular velocity virtual control quantity ω _ref and the temperature control load output power virtual control quantity P _{TCL, ref} in turn, then the control input U _i of each temperature control load aggregate can be obtained. (t).

The system control flow is shown in Figure 1.

Finally, through the algorithm simulation verification analysis, the accuracy of the improved heterogeneous TCL bilinear aggregation model and the regulation performance of the distributed hierarchical multi-objective coordinated control are tested.

Description of drawings

Fig. 1 is the system control flow chart;

Figure 2 shows the coupled dynamic characteristics of air and indoor medium temperature;

Figure 3 shows the second-order heterogeneous ETP model of monomeric TCL;

Figure 4 shows the dynamic process of TCL finite difference discretization;

Figure 5 shows the comparison of the aggregation power between the heterogeneous TCL bilinear aggregation model and the original TCL bilinear aggregation model;

Figure 6 is a comparison of the aggregation power between the improved heterogeneous TCL bilinear aggregation model and the heterogeneous TCL bilinear aggregation model;

Figure 7 is a structural block diagram of an IEEE 9 node system simulation example;

Figure 8 shows the system power generation plan and load forecast;

Figure 9 shows the power shortage that TCL needs to absorb;

Figure 10 shows the system power angle fluctuation;

Figure 11 shows the system frequency fluctuation;

Figure 12 shows the total power tracking error in each control mode: (a) centralized control, (b) distributed control, (c) decentralized control;

Figure 13 shows the output power coefficient of the aggregate under each control mode: (a) centralized control, (b) distributed control, (c) decentralized control;

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings:

1 System Model

1.1 Improved heterogeneous TCL bilinear aggregation model

1.1.1 Thermal dynamic model of a single temperature control load

Most of the existing TCL aggregation modeling methods are based on first-order thermodynamic ordinary differential equations to simulate its thermal dynamic process, ignoring the second-order dynamic effects caused by the material temperature _Tm . This simplification often leads to the construction of a model that cannot accurately describe the real TCL dynamic process in transient responses such as those caused by temperature set-point changes. Figure 2 shows the evolution of the air temperature T _a and material temperature T _m of the air conditioning load when the temperature set point is changed from 25°C to 27°C at time t=1 (hr). It can be seen that it takes _a longer time for Ta to increase from 25°C to 27°C during the time period [t ₁ , t ₂ ] compared to the time period [t ₃ , t ₄ ]. This is due to the fact that the lower mass temperature _Tm significantly _slows the increase in air temperature Ta during [ _t1 , _t2 ].

Without considering T _m , the time derivative of T _a depends only on T _a itself, which ignores the effects shown in Fig. 2. In fact, only the first-order model cannot accurately describe the entire trajectory of _Ta as presented in Figure 2. Therefore, detailed consideration of the coupled thermodynamic properties of _Tm and _Ta is necessary to obtain high-quality aggregation models that can accurately describe the transient and steady-state aggregation responses of TCL populations.

In the following, the second-order differential equation Equivalent Thermal Parameter (ETP) model is used to describe the monomer TCL, and the indoor temperature and the indoor material temperature change are the two state variables observed in the ETP model, so it is also referred to as different The mass model mainly considers the heat exchange process between the outdoor air and the indoor air of the house, as well as the heat exchange process between the indoor material and the indoor air, the heat loss in the exchange process, and the heat storage process is described by heat capacity and thermal resistance modeling, dual medium The specific differential equation of the model is:

T_a(t)为室内空气温度，T_m(t)为室内物质温度，T_∝为室外环境温度，P为单个TCL操作电功率，T_max、T_min为温度限制的上下限，T_set为温度设定值，Δ_db为温度死区，C_a为室内空气热容，C_m为室内物质热容，R_a为室内空气热阻，R_m为室内物质热阻。其热力学等值模型如图3所示，其中，UA_air为室内空气热损失系数，UA_mass为室内物质热损失系数，R_a＝1/UA_air，R_m＝1/UA_mass。in,

T _a (t) is the indoor air temperature, T _m (t) is the indoor material temperature, T _∝ is the outdoor ambient temperature, P is the operating electric power of a single TCL, T _max and T _min are the upper and lower limits of the temperature limit, and T _set is the temperature The set value, _Δdb is the temperature dead zone, C _a is the indoor air heat capacity, C _m is the indoor material heat capacity, R _a is the indoor air thermal resistance, and R _m is the indoor material thermal resistance. The thermodynamic equivalent model is shown in Figure 3, where UA _air is the heat loss coefficient of indoor air, UA _mass is the heat loss coefficient of indoor material, R _a =1/UA _air , R _m =1/UA _mass .

The above thermocouple model aggregated power P _r is:

In the formula, η _i is the efficiency coefficient of load i.

1.1.2 Heterogeneous TCL Bilinear Aggregation Model

Although the centralized TCL represented by the ETP model can accurately describe the power consumption characteristics of the TCL cluster, its model has both continuous state variables (temperature) and discrete on/off state variables, which are difficult to directly use in control design. ; and if each TCL is expressed as a set of independent differential equations, the model will face the dilemma of dimension disaster when it is applied to grid-level load dynamic response.

A control-oriented bilinear aggregation model of heterogeneous TCL is derived below. In this model, it is assumed that all temperature control loads are in a limited temperature range [ _TL , _TH ], the entire temperature range is discretized into small segments of equal width, and each temperature segment contains "on/off" two There are Q temperature state intervals in total, and each TCL belongs to a certain discrete temperature state interval, as shown in FIG. 4 .

A heterogeneous TCL bilinear aggregation model is constructed, as shown in equation (3).

为控制输入；P_r(t)为聚合TCL的总输出功率；

为1×Q阶输出矩阵；A为Q×Q阶系统矩阵，描述为In the formula: X(t)=[x ₁ (t), x ₂ (t), …, x _Q (t)] ^T is the Q×1-order state variable matrix, representing the temperature range in each temperature interval after the finite difference discretization The number of TCLs;

is the control input; P _r (t) is the total output power of the aggregated TCL;

is a 1×Q-order output matrix; A is a Q×Q-order system matrix, described as

where:

respectively represent the load transfer rate of the TCL in the "on/off" state.

Matrix B is a Q×Q order bilinear matrix, which has the same structure as matrix A. Set α _on/off in matrix A to -1 to obtain matrix B, that is, B=A(-1,-1 ).

及初始物质温度T_m0下的平均负荷传递率，即：In order to reduce the iterative calculation of the matrix in the operation, the load transfer rate of each temperature interval is approximated to the desired temperature set value

and the average load transfer rate at the initial material temperature T _m0 , namely:

将是一个常数矩阵，可大大减少计数据计算及存储量，提高模型运算速度。but

It will be a constant matrix, which can greatly reduce the calculation and storage of count data, and improve the speed of model operation.

1.1.3 Improved heterogeneous TCL bilinear aggregation model

较大时，如果依然用

下的平均负荷传递率

代替α_on/off，就会造成双线性聚合模型失效，长时间运行后会出现显著偏差。The heterogeneous TCL bilinear aggregation model is only suitable for scenarios with small changes in load temperature setpoints. When T _set (t) deviates from the desired set value

larger, if you still use

Average load transfer rate under

Replacing α _on/off will cause the bilinear aggregation model to fail, with significant deviations after long runs.

始终不变，不能反映实时的内部温度T_a(t)，下面，利用实时物质温度T_m(t)和实时温度设定值T_set(t)来改进对α_on/off的估计：During a control cycle,

It is always unchanged and cannot reflect the real-time internal temperature T _a (t). Below, the real-time material temperature T _m (t) and the real-time temperature set value T _set (t) are used to improve the estimation of α _on/off :

From this, the improved heterogeneous TCL bilinear aggregation model can be derived:

控制输入u₁(t)、u₂(t)、u₃(t)分别体现了温度设定值变化对温控负荷的实时影响、温度设定值变化对温控负荷的累积影响以及异质性对温控负荷的影响。其中T_m(t)在现实中不易得到，可转由用T_set(t)代替T_a(t)根据式(1)估算获取。In the formula,

The control inputs u ₁ (t), u ₂ (t), and u ₃ (t) respectively reflect the real-time effect of temperature setpoint changes on the temperature control load, the cumulative effect of temperature setpoint changes on the temperature control load, and the heterogeneity. The effect of temperature on the temperature control load. Among them, T _m (t) is not easy to obtain in reality, and can be obtained by replacing T _a (t) with T _set (t) according to formula (1).

1.2 Equivalent model of generator inertial center system

It is assumed that there are S generator sets in the system, and the generators all adopt the classical second-order model that ignores the saliency effect. It is assumed that the generator excitation system is strong enough to keep the transient electromotive force amplitude after the generator transient reactance unchanged in the dynamic process. , the generator model is obtained as follows:

In the formula, δ _i is the power angle of generator i, ω _i , ω _i,0 are the angular velocity and standard angular velocity of generator i respectively, T _j,i is the inertia time constant of generator i, P _m,i , P _{e, i} are the mechanical power and electromagnetic power of generator i, respectively, and D _i is the constant damping coefficient of generator i.

When the above model is used to calculate the frequency control strategy of a large-scale power grid, it is found that with the increase of the system scale, the time-consuming solution increases rapidly, and even unsolvable situations may occur. This is because the frequency control model proposed in this section contains the constraints of large-scale differential-algebraic equations, time-varying variables and time-invariant variables are coupled with each other, and the complexity of the model solution increases rapidly as the number of system nodes increases. Therefore, it is necessary to further analyze the problem to improve the solution speed by improving the built model.

The electrical angular velocity of each generator can be regarded as approximately the same, and the generator in the power grid is equivalent to a generator under the inertial center system, which simplifies the transient power angle stability constraint and establishes an equivalent generator model under the inertial center system. ,which is:

δ_s、ω_s、ω_s，0、T_j，s、D_s分别为惯性中心系统的等值功角、等值角速度、标准角速度、等值时间常数以及等值阻尼系数；P_m，s为系统总机械功率，P_G，i为发电机组i的发电功率，P_e，s为系统总电磁功率，P_TCL为温控负荷总输出功率，P_L为系统刚性负荷，P_new为系统分布式电源发电功率，P_AGC为AGC机组增加的发电功率。in,

δ _s , ω _s , ω _s,0 , T _{j, s} , D _s are the equivalent power angle, equivalent angular velocity, standard angular velocity, equivalent time constant and equivalent damping coefficient of the inertial center system, respectively; P _{m, s} is the total mechanical power of the system, P _{G, i} is the generated power of the generator set i, P _{e, s} is the total electromagnetic power of the system, P _TCL is the total output power of the temperature-controlled load, P _L is the system rigid load, and P _new is the system distribution P _AGC is the power generated by the AGC unit.

The equivalent power angle and frequency of the inertial center system can be considered as the power angle and frequency of the entire power system. As long as the stability of the inertial center system parameters is maintained, the entire regional power grid system can be considered to be in a stable operation state.

2 Distributed hierarchical multi-objective coordinated control design

2.1 Problem description

In the layered architecture based on load aggregators, load aggregators, as the intermediary agency that coordinates a large number of small and medium-sized users and grid control centers, can be traditional power distribution companies, government entities, or the load management centers of the grid companies themselves, or they can be A third-party organization representing a single type or multiple types of loads, the common ground is that a large number of power end users are brought together to participate in grid scheduling, and strive to achieve the established goals of grid companies, load aggregators and power end users. The load aggregator obtains the controllability and scheduling response capacity of each load cluster from the managed load group, thereby establishing the response model of the entire load group and realizing its own distributed autonomy function.

The main purpose of designing the coordination controller in this paper is to enable the load aggregator to evenly distribute the output power of each TCL aggregate, and to regulate each TCL aggregate to accurately track the issued power quota to ensure system power balance, maintain power angle and Frequency stabilization, that is, this study is a hierarchical multi-objective coordinated control problem.

2.2 Consistent pinning control strategy

According to the multi-agent consistency pinning control method, the control state information of each agent can be expressed as:

u _i (t+Δt)=f _i [h _i0 (t)u ₀ (t), h _i1 (t)u ₁ (t), ..., h _iN (t)u _N (t)] (11)

In the formula: u _i (t) represents the control information of the i-th main body at time t; u ₀ (t) represents the control information of the control center at time t; h _ij represents the i-th main body and the j-th main body If the communication between the i-th main body and the j-th main body can be communicated, then h _ij =1, otherwise h _ij =0. Furthermore, if the i-th main body can communicate with the control center, then h _0i =1, otherwise h _0i =0. h _ii =1 is suitable for any subject, which means that all subjects can obtain their own information.

The time-varying communication coefficients can be represented by a communication topology matrix:

According to its own stored historical data and forecast rolling model, the dispatch center can directly obtain the daily power generation and consumption plan values of units and load agents for several hours in the future, the new energy power injection value in the system and the load forecast value of the system. The total power deficit of the system can be expressed as:

ΔP=P _L +P _TCL -P _AGC -P _m,s -P _new (13)

To balance the output power of each load aggregate, you can set:

In the formula, Pi is the power consumption of the _ith TCL aggregate, Pi _,max is the maximum power of the ith TCL aggregate, and λ is the output power coefficient of the TCL aggregate.

The control signal λ ₀ (t) of the control center can be constructed as follows based on the PI control idea:

The distributed consistency control algorithm of each TCL aggregate can be described as:

In the formula, λ _i (t) is the time-varying output power coefficient, and _{lk, i} > 0 is the control gain of the load aggregate, which is taken as:

As long as the system power deficit ΔP≠0, λ ₀ (t) and λ _i (t) will iterate until the system power is balanced, and the values of each λ _i (t) tend to be consistent during the iterative calculation process, ensuring that every Each aggregate is assigned an output power tracking value corresponding to its own maximum power ratio.

2.3 Reverse push control strategy

Simultaneous equations (8) and (10) can be obtained:

According to the inverse control theory, there are:

1) Let e _δ =δ _{s, ref} −δ _s , where e _δ is the system phase error variable, and δ _{s, ref} is the reference phase. According to the inverse control principle, derivation of e _δ , we can get:

Design the first virtual control quantity ω _{s, ref} is as follows:

ω _{s, ref} = ω _{s, 0} -k _δ e _δ (20)

where k _δ is an adjustable phase control parameter greater than 0. Substitute equation (20) into equation (19) to get:

2) Let e _ω =ω _{s, ref} -ω _s , where e _ω is the system angular velocity error variable, and ω _{s, ref} is the reference angular frequency. Derivative with respect to e _ω , we can get:

Design the second virtual control quantity P _{TCL, ref} is as follows:

where k _ω is an adjustable angular velocity control parameter greater than 0. Substitute equation (23) into equation (22) to get:

3) by

The tracking reference power of each temperature-controlled load aggregate can be calculated as:

Then e _P,i =P _i,ref -P _i , where e _P,i is the system power error variable corresponding to the ith TCL, P _i,ref is the tracking reference power allocated to the ith TCL, P _i =C _i X _i (t) is the aggregated power of the i-th aggregated TCL. Differentiating e _{P, i} , we can get:

The actual control quantity U _i (t) is designed as follows:

In the formula, k _{P, i} are adjustable power control parameters greater than 0. Substitute equation (28) into equation (27) to get:

3 Model verification and algorithm implementation and analysis

3.1 Simulation Analysis of Improved Heterogeneous TCL Bilinear Aggregation Model

In order to check the accuracy of the optimized dual-media bilinear model, 1000 TCLs were selected for simulation (C _a =7.5, C _m =2.5, R _a =2, R _m =1, P = 7, η _r =2.5, δ _db =1), and the results are shown in FIGS. 5 to 6 .

It can be seen that under the same parameters, compared with the original bilinear aggregation model that only considers the air medium in Fig. 5, the thermal resistance properties of the indoor material make the power fluctuation of the heterogeneous bilinear model when the temperature set value changes. The period is longer and the fluctuation amplitude is larger; in Figure 6, the optimization of equation (7) improves the cumulative error problem of the original bilinear model, so that the long-term control accuracy of the improved heterogeneous TCL bilinear model is obtained. Greatly improved, providing the feasibility for TCL clusters to participate in system regulation.

3.2 Distributed hierarchical multi-objective coordinated control simulation analysis

The above distributed hierarchical multi-objective coordinated control strategy is verified in a small power system, in which there are 3 generating units and 1 demand response load aggregator in the regional grid, and the units and load agents are placed in an improved IEEE 9 node system, And the load agents are all considered as refrigeration-type air-conditioning equipment. The system also has a wind farm and a photovoltaic power station. The physical connection and communication connection between the unit and the load agent are shown in Figure 7.

The parameters of each unit are as follows:

Table 1 Unit parameters

Assuming 8 load clusters, a total of 25,418 cooling-type temperature control devices participate in the regulation of the power system. The parameters of the temperature-controlled load polymer are selected from the value range in the table below according to the average distribution.

Table 2 Typical TCL polymer parameter range

The power generation and consumption load data in the next 2 hours at a certain moment in the system can be taken as shown in Figure 8 below. It can be seen that the first half of the system is the peak load period within the 120min period, and the second half of the new energy power generation increases sharply. In the actual operation system, the AGC adjustment capacity is generally selected as 2% to 5% of the maximum load of the system, in which the large system takes the smaller percentage value, and the small system takes the larger percentage value. In this example, the maximum load of the system is taken as 1, and the AGC economic dispatch capacity is taken as 4%, that is, the AGC economic dispatch capacity is ±0.04. Then the system power shortage after filtering out the AGC economic dispatch is shown in Figure 9 below.

In order to better test the control effect of the distributed hierarchical multi-objective coordinated control strategy, a comparative simulation experiment was designed to compare it with the centralized control and the decentralized control. Among them, centralized control can be constructed by setting h _0i =1, h _ii =1 and h _ij =0 (i≠j) in the communication matrix h; decentralized control is constructed based on the idea of decentralized hierarchical control. The simulation results are shown in Figure 10 to Figure 13.

It can be seen from Figure 10 and Figure 11 that the multi-objective coordinated control strategy based on the inverse control structure can correctly control each TCL aggregate to respond to the regulation target in the centralized control mode and the distributed control mode, and maintain the system stability. Further observation in conjunction with Fig. 12 shows that the centralized control has the best control effect on the power angle and frequency of the system, and the load tracking error is also the smallest; the distributed control control effect is second, but the fluctuation of the power angle and frequency is still within the allowable range of the system; In contrast, distributed control has the worst stabilization effect on the power angle and frequency of the system. Under the current system parameters, the system frequency fluctuation is close to the allowable critical value during the control process, and the control of the power angle has a steady-state error. It can be seen from Figure 13 that the distributed control and the centralized control use a consistent control algorithm, and the output ratio of each polymer is relatively consistent, while under the distributed control, each polymer responds freely, and the output ratio shows irregular changes. Combining with Fig. 12, it is further observed that the power and tracking error fluctuation of distributed control and centralized control are smaller in steady state, and the static stability is better, while the power and tracking error of distributed control fluctuate larger in steady state, showing Some degree of static instability. The reasons for the above control differences are analyzed as follows:

1. The inverse control is a control strategy based on the Lyapunov stability principle, which has strong robustness to the control object; the consistency control coordinates the output ratio of each aggregate to be consistent, which can greatly improve the system performance. computability;

2. Centralized control Since each aggregate can directly receive system control information from the control center, no intermediate transmission link is required, which can greatly reduce errors in the control process and show excellent tracking performance. But this is based on the high-speed and effective communication between the control center and each aggregate. If there is a communication line failure, the system will lose the control capacity of this aggregate, which will affect the communication reliability, communication transmission synchronization and communication quality of the system. put forward extremely high requirements;

3. Distributed control within the allowable range, by sacrificing a certain control accuracy, in exchange for great control robustness. Distributed control does not require each aggregate to get in touch with the control center, and the communication between the upper and lower layers of most aggregates and the control center is changed to local communication between aggregates, which greatly reduces the communication requirements of the control system. Even if the control effect under the high sparse communication matrix taken in this example is satisfactory, and if the communication connection between aggregate 5 and aggregate 6 is interrupted, aggregate 6 can still be obtained through aggregate 7 Control information participates in system regulation.

4. Distributed control Because there is no unified deployment center, it is difficult for each polymer to coordinate with each other, and it is prone to insufficient or excessive response, and even the problem of mutual cancellation of control effects, and there may be static fluctuations in distributed control; Aggregates are free to respond and it will be more difficult to obtain the state of all aggregates in the control process. These factors are not conducive to the stability control of the system.

To sum up, this paper expands the homogeneous bilinear aggregation model of TCL into an improved heterogeneous bilinear aggregation model, and optimizes the cumulative error existing in the bilinear aggregation model. The generator is equivalent to a generator under the inertial center system. A distributed hierarchical multi-objective coordinated control method based on consistency control and reverse thrust control is proposed to coordinate multiple temperature-controlled load aggregates to fully participate in power system regulation. . The research has verified that the optimization of the improved heterogeneous bilinear aggregation model improves the accuracy of the model description; the distributed hierarchical multi-objective coordinated control method can effectively control the system state under different control modes, making the power angle, frequency and power Multiple control objectives such as the control system can quickly track their respective reference values, which shows the correctness and good robustness of the coordinated control strategy. Distributed hierarchical multi-objective coordinated control can greatly reduce the communication requirements of the control system. Robustness and computability are superior to traditional centralized and decentralized control.

The above description is only the preferred embodiment of the present invention, but it should not be construed as a limitation on the scope of the patent of the present invention. Any equivalent replacement, modification, improvement, etc. made within the principle and spirit of the present invention shall include within the protection scope of the present invention.

Claims (3)

1. An improved heterogeneous temperature control load bilinear aggregation model and its distributed hierarchical multi-objective coordinated control method, including an improved heterogeneous temperature control load bilinear aggregation model constructed based on the original homogeneous temperature control load bilinear aggregation model A distributed hierarchical multi-objective coordinated control method for coordinating multiple temperature-controlled load aggregates is designed based on consistent control and inverse control. 2. The improved heterogeneous temperature control load aggregation model and its distributed hierarchical multi-objective coordinated control method as claimed in claim 1, characterized in that: The bilinear aggregation model of the original homogeneous temperature-controlled load derived from the first-order thermodynamic parameter model is extended by using the equivalent thermodynamic parameter model of the second-order differential equation of the temperature-controlled load, and the load transfer rate of each temperature interval is Approximate the desired temperature setpoint

and the average transport rate at the initial material temperature T _m0 , the bilinear polymerization model of heterogeneous temperature-controlled load is obtained:

In the formula: X(t)=[x ₁ (t), x ₂ (t), …, x _Q (t)] ^T is the Q×1-order state variable matrix, representing the temperature range in each temperature interval after the finite difference discretization The number of temperature control loads, Q is the total number of temperature ranges;

is the control input, T _set (t) is the temperature set value of the temperature control load; P _r (t) is the total output power of the aggregated temperature control load;

is a 1×Q-order output matrix, P is the operating electric power of a single temperature-controlled load; A is a Q×Q-order system matrix, described as:

in,

respectively represent the average load transfer rate of the temperature control load in the “on/off” state; C _a is the indoor air heat capacity, C _m is the indoor material heat capacity, R _a is the indoor air thermal resistance, R _m is the indoor material thermal resistance, T _∞ is the outdoor ambient temperature, ΔT is the discrete step size; Matrix B is a Q×Q order bilinear matrix, which has the same structure as matrix A. Set α _on/off in matrix A to -1 to obtain matrix B, that is, B=A(-1,-1 ); Then, in the process of derivation of the bilinear aggregation model,

Optimize the cumulative error problem caused by the approximation, and replace it with the real-time temperature set value T _set (t)

middle

Replace with real-time material temperature T _m (t)

T _m0 in , the final derivation of the improved heterogeneous temperature-controlled load bilinear aggregation model is:

In the formula,

The control inputs u ₁ (t), u ₂ (t), and u ₃ (t) respectively reflect the real-time effect of temperature setpoint changes on the temperature control load, the cumulative effect of temperature setpoint changes on the temperature control load, and the heterogeneity. The effect of temperature on the temperature control load. 3. the improved heterogeneous temperature control load bilinear aggregation model and its distributed hierarchical multi-objective coordinated control method as claimed in claim 1, it is characterized in that: According to the principles of consistency control and reverse push control, a distributed hierarchical multi-objective coordinated control method for temperature-controlled load multi-polymer control is designed; in which, the consistency control is applied to the aggregator level to balance each temperature-controlled load. The output of the aggregate, the communication topology matrix of consistency control is expressed as:

In the formula, h _ij represents the communication link between the i-th aggregate and the j-th aggregate, if the i-th aggregate and the j-th aggregate can communicate, then h _ij =1, otherwise h _ij = 0; In addition, if the i-th aggregate can communicate with the control center, then h _0i =1, otherwise h _0i =0; h _ii =1 is suitable for any aggregate, indicating that all aggregates can obtain their own information ; In order to make each temperature-controlled load aggregate output power in the same proportion, set:

In the formula, Pi is the output power of the _ith temperature-controlled load aggregate, Pi _,max is the maximum power of the ith temperature-controlled load aggregate, and λ is the output power coefficient of the temperature-controlled load aggregate; The control signal λ ₀ (t) of the control center is constructed as follows based on the PI control idea:

In the formula, ΔP is the system power deficit; The consistency control algorithm of N temperature-controlled load aggregates can be described as:

In the formula, λ _i (t) is the time-varying output power coefficient, and _{lk, i} > 0 is the control gain of the load aggregate, which is taken as:

As long as ΔP≠0, λ ₀ (t) and λ _i (t) will iterate until the system power is balanced, and the values of each λ _i (t) tend to be consistent during the iterative calculation process, ensuring that each aggregate It is assigned to the output power tracking value corresponding to its own maximum power ratio; The reverse push control is applied to the temperature-controlled load aggregate level, and is used to control the output power of the temperature-controlled load to track the output quota allocated by the load aggregator. The mathematical model for constructing the power system and the temperature-controlled load is:

where _δs is the equivalent power angle of the inertial center system, ωs is the equivalent angular _velocity of the inertial center system, ωs _{, 0} is the standard angular velocity of the inertial center system, _{Tj, s} is the equivalent time constant of the inertial center system, D _s is the equivalent damping coefficient of the inertial center system; P _{m, s} is the total mechanical power of the system, P _{e, s} is the total electromagnetic power of the system, P _TCL is the total output power of the temperature control load, _PL is the rigid load of the system, and P _new is the system Distributed power generation power, P _AGC is the increased power generation of AGC units; X _i (t), A _i , B _i , U _i (t) represent the state variable matrix, system matrix, dual Linear matrices and control inputs; According to the reverse push control principle, the given power angle reference value is δ _ref , and after deriving the angular velocity virtual control quantity ω _ref and the temperature control load output power virtual control quantity P _TCL,ref in turn, the control input of each temperature control load aggregate can be obtained. U _i (t) is:

In the formula, e _{P, i} is the system power error variable corresponding to the ith TCL, and k _{P, i} is an adjustable power control parameter greater than 0.

Images