WO2022127652A1

WO2022127652A1 - Air conditioning aggregation control method, system, and regulation apparatus

Info

Publication number: WO2022127652A1
Application number: PCT/CN2021/136180
Authority: WO
Inventors: 胡秦然; 陈心宜; 赵原; 张玮椿; 汤景然; 全相军; 窦晓波; 吴在军
Original assignee: 东南大学
Priority date: 2020-12-17
Filing date: 2021-12-07
Publication date: 2022-06-23
Also published as: CN112636360A

Abstract

Provided are an air conditioning aggregation control method, system, and regulation apparatus, belonging to the field of electric power system demand response load aggregation control; according to a load aggregation control objective, on the basis of historical feedback from users in response to an adjustment instruction, taking, in the historical feedback, as a reference the smaller of the higher probability estimate I of receiving an adjustment instruction, the smaller variance of response behavior II, and the smaller number of historical times nt,i of being sent an instruction, setting the sort index value vt,i for the user, and selecting m users in order from largest to smallest by sorting index value vt,ito send an adjustment instruction until III; if III, then letting m = n and P1,..., Pn denote the air conditioning power of the user, respectively, n being the total number of users. The invention achieves a more desirable secondary frequency modulation effect while avoiding blindly sending an adjustment instruction to the user, reducing the number of users sending instructions and the number of users opting out of responses, reducing load aggregator cost and user fatigue due to frequent adjustment instructions.

Description

A kind of air conditioning polymerization control method, system and regulating device

technical field

The invention belongs to the field of power system demand response load aggregation control, and more particularly relates to an air conditioning aggregation control method, system and regulation device.

Background technique

With the large-scale access of renewable energy sources such as photovoltaics and wind power, their intermittency and volatility have brought great challenges to the stable operation of the power grid, increasing the demand for flexible resources in the power system. At the same time, due to the gradual replacement of a large number of traditional generator sets by new energy sources, it is very difficult to maintain real-time power balance and frequency stability of the system only by relying on power supply side resources to track load changes in traditional mode. With the rapid development of advanced measurement infrastructure and information and communication technology, demand-side resources have great advantages in balancing power supply and demand and promoting renewable energy consumption due to their considerable adjustment potential and flexible control methods. However, how to Reliable and efficient integration of demand-side resources is a necessary prerequisite for unlocking their controllable potential.

In a demand response contract, the contracted users often have the right to opt out of a response event or terminate the contract at any time due to personal reasons or environmental factors, and their behavior is easily affected by many factors, so that in each demand response event , the user's response to the regulation command issued by the load aggregator can be regarded as a random variable, and its uncertainty and randomness make it difficult for demand-side resources to be used as a reliable aggregated resource to provide auxiliary services such as frequency and peak regulation for the power grid.

The current research on the use of load aggregation for auxiliary services either focuses on achieving faster adjustment command response speed, or focuses on achieving higher economic benefits. However, because the uncertainty of user response behavior itself is not taken into account, load aggregation is reduced. The reliability and precision of the control method.

SUMMARY OF THE INVENTION

In view of the above defects and improvement needs of the prior art, the present invention proposes an air conditioner aggregation control method, system, and control device, which aim to overcome the uncertainty of the user's response to the adjustment command behavior through the online learning method, and realize the improvement of the air conditioner load. Accurate aggregation control achieves a more reliable secondary frequency regulation effect, while reducing the number of users who send commands and the number of users who opt out of the response, thereby reducing the cost of load aggregators and the fatigue of users due to frequent adjustment commands.

In order to achieve the above-mentioned purpose, the present invention adopts and provides a kind of air-conditioning polymerization control method and system, comprising the following steps:

An air conditioning polymerization control method, comprising the following steps:

According to the load aggregation control objective, based on the historical feedback of the user's response to the adjustment instruction, the probability of accepting the adjustment instruction in the historical feedback is estimated

Higher, response behavior variance

It is relatively small and the number of historical times n _t,i of sent instructions is less as a reference, set the user's sorting index value v _t,i , and select m users in descending order of the sorting index value v _t,i to send adjustment instructions ,until

like

Then let m=n, P ₁ , . . . , P _n respectively represent the air-conditioning power of users, and n is the total number of users.

Further, the sorting index value of the user

Further, the load aggregation control objective is to minimize the deviation of the actual power regulation amount D _agg,t compared to the secondary frequency regulation target D _tar,t .

Further, at time t, the secondary frequency modulation power adjustment target D _tar,t is the smaller value between the adjusted power value required to restore the system frequency air conditioning and the actual available adjustment reserve.

Further, the adjusted power value required to restore the system frequency air conditioner is the product of the secondary frequency modulation coefficient k _S and the system frequency deviation |Δf(t)|.

Further, the actual available adjustment reserve in real time

η represents the cooling coefficient of the air conditioner, s _i (t) represents the switching state of the _ith air conditioner at time t, and Qi represents the thermal power of the ith air conditioner.

Further, after an adjustment instruction is sent to the users selected according to the index value sorting according to the load aggregation control target, the action information of the user in response to the adjustment instruction is collected and used to optimize the sorting index value selected by the next user.

An air-conditioning aggregation control system, comprising: a load operation monitoring module, a system frequency sensing module, a power grid dispatch center, a load control center module, an air-conditioning local control module and a communication module;

The load operation monitoring module is used to estimate in real time the actual available adjustment reserve after the N air conditioner loads are aggregated;

The system frequency sensing module is used to sense the system frequency change in real time;

The power grid dispatch center, based on the system frequency sensing module sensing the system frequency change and the actual available regulation reserve of the load operation monitoring module, formulates the secondary frequency regulation power regulation target;

The load control center module is used to formulate a strategy selected by the user after receiving the secondary frequency regulation power adjustment target value assigned by the power grid dispatch center, and send the adjustment command to the user, and collect the user's feedback on the command at the same time;

The local control module of the air conditioner performs corresponding on-off operations on the air conditioner according to the user's response to the adjustment command of the load control center.

The communication module is used to realize the information exchange between the grid dispatching center and the load control center, the load control center and the local control module of the air conditioner.

A control device includes a processor, and the processor is used for executing a program to implement the above-mentioned method or the above-mentioned system flow.

Compared with the prior art, the above technical solutions conceived by the present invention can achieve the following beneficial effects:

The invention improves the reliability of load aggregation, realizes a more ideal secondary frequency modulation effect, and at the same time avoids aggregators blindly sending adjustment instructions to users, reduces the number of users who send instructions and the number of users who choose to quit responding, thereby reducing Load aggregator cost and user fatigue due to frequent adjustment commands.

Description of drawings

Fig. 1 is a kind of air-conditioning polymerization control method and system flow chart provided by the present invention;

Figure 2 is a block diagram of the implementation architecture of air-conditioning load aggregation to participate in secondary frequency regulation;

Figure 3 is an improved IEEE RTS 24 node system diagram;

Fig. 4 is the frequency diagram after one frequency modulation of system disturbance;

Fig. 5 is the method provided by the present invention and the traditional method frequency modulation result comparison chart;

6 is a comparison diagram of the actual aggregated power between the method provided by the present invention and the traditional method;

FIG. 7 is a comparison diagram of the relative target deviation of the actual aggregated power between the method provided by the present invention and the traditional method;

Fig. 8 is the method that the present invention provides and the traditional method to select the user number comparison diagram;

FIG. 9 is a comparison diagram of the number of user exits between the method provided by the present invention and the traditional method.

Detailed ways

In order to make the purpose, technical solutions and beneficial effects of the present invention more clearly displayed, the present invention will be further illustrated below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. After reading the present invention, those skilled in the art make modifications to various equivalent forms of the present invention within the scope defined by the appended claims of the present application.

Example 1:

S1. Establish a heat transfer model of a single air conditioner, and estimate in real time the actual available adjustment reserve P _AC after the loads of N air conditioners are aggregated;

Specifically, the heat transfer model of a single air conditioner is:

The discrete form of this formula is:

Among them, C and R are the thermal capacity and thermal resistance, respectively, T _in (t) and T _out are the indoor and outdoor temperatures, Q is the thermal power of the air conditioner, s (t) represents the on and off states of the air conditioner, and the air conditioner is in the cooling state. Taking the mode as an example, its determination is based on:

Among them, T _max and T _min represent the upper and lower bounds of the indoor temperature, respectively.

The real-time actual available adjustment reserve P _AC (t) after the aggregation of N air-conditioning loads is estimated as:

Among them, η represents the cooling coefficient of the air conditioner, s _i (t) represents the switching state of the _ith air conditioner at time t, and Qi represents the thermal power of the ith air conditioner.

S2. For the system frequency deviation, calculate the secondary frequency modulation power adjustment target according to the secondary frequency modulation coefficient and the actual available adjustment reserve P _AC in the system;

Specifically, the system frequency deviation is:

|Δf(t)|=|f(t)-f _N |

Among them, f _N is the rated frequency of the grid.

The secondary frequency modulation coefficient is used to simulate the droop characteristics of the traditional power generation side, which can be calculated by the following formula:

where max(P _AC ) is the maximum regulation reserve available, and |Δf _mSFR | is the maximum frequency deviation that the system can withstand during secondary frequency modulation.

Therefore, the target D _{tar,t of the load side participating in the secondary frequency regulation power regulation at time t} should take the smaller value between the power value that needs to be regulated to restore the system frequency and the actual available regulation reserve P _AC (t), that is, :

D _tar,t =min(k _S |Δf(t)|,P _AC (t))

S3. In order to achieve the secondary frequency modulation power adjustment target D _tar,t in S2, the load aggregator needs to select among N users

Each user sends an adjustment command. If the user accepts the adjustment command, the air conditioner will be temporarily turned off/on, but not all commands will be responded to at the same time. Even if the demand response contract is signed or the user is invited in advance, there will be Due to environmental or personal factors, some users will not respond to the adjustment command as expected but choose to withdraw, so that the actual power adjustment amount D _agg,t cannot match the secondary frequency modulation power adjustment target D _tar,t , and the more users who choose to withdraw, The greater the deviation between D _agg,t and D _tar,t . Therefore, a behavioral model considering the uncertainty of the user's response to the secondary FM load regulation command is established;

Specifically, the response of user i to the adjustment command can be regarded as a random variable obeying the Bernoulli distribution, that is, the user has a probability of p _i to agree to accept the adjustment command, corresponding to the air-conditioning load action; there are (1-p _i ) Probability to opt out of the response, the corresponding air conditioner does not operate.

The expected value is:

E(X _i )= _pi

The variance is:

σ _i ² = _pi (1- _pi )

S4. Based on the uncertain user response behavior model in S3, in order to use the aggregated load as a reliable resource to achieve accurate secondary frequency regulation, the goal of the load aggregator is to reduce the actual power adjustment amount D _{agg as much as possible, t} is the deviation of the secondary frequency modulation power regulation target D _tar,t , so it describes the load aggregation control target;

Specifically, the load aggregation control objectives are:

min E(D _agg,t -D _tar,t ) ²

It should be noted that due to the randomness of the user's behavior, the actual power adjustment amount D _agg,t is also random, so the above goal is to minimize the expected value of the squared difference between D _agg,t and D _tar,t . express.

S5. In order to achieve the load aggregation control goal in S4, the user's historical response feedback is learned online based on the framework of "risk avoidance dobby", so as to help make the next user selection decision. After sending adjustment instructions to the user, further Observe user feedback, and so on to form a closed-loop process.

Specifically, the aggregator forms the existing cognition of the user according to the historical feedback of the user's response to the adjustment instruction, including an estimate of the probability of the user accepting the adjustment instruction

(The number of times of historical acceptance of adjustment / the number of times of instructions sent) and variance

And the historical number of sent instructions n _t,i ; Combined with the above knowledge, before selecting users, the user is prioritized based on the framework of "risk avoidance dobby". Under this framework, on the one hand, the aggregator The existing knowledge of users should be fully utilized, and users with higher historical response power expectations and smaller variance should be selected. On the other hand, it is also necessary to explore the behaviors of more users to form a broader understanding of different users. The trade-off with exploration can calculate the sorting index value v _t,i of each user; then select m users according to the index value from large to small to send adjustment instructions, until

like

Then let m=n, P ₁ , . . . , P _n respectively represent the air-conditioning power of the user, and n is the total number of users; after the command is sent, the user's cognition is updated online in real time according to the user's response result, as the next user basis for selection.

_The above process can be further expressed as: the first step, input parameters: ρ ₁ , ρ ₂ , P ₁ ,...,P _n , D _tar,t _; The “machine” learning framework calculates the weights of the risk aversion item (tends to select users with a small historical response variance) and the exploration item (tends to select users with fewer historical selections) when the user selects the priority;

In the second step, for any user i∈[N], initialize the adjusted probability estimate

variance estimate

And the number of times n _t,i of historically sent adjustment instructions;

The third step, at any time t∈[T]:

Calculate the sort index value for each user:

Sort users in descending order of the size of the index value v _t,i :

Select m users in sequence until:

like

Then let m=n;

Output the set of users who choose to send FM commands: S _t ={ε ₁ ,...,ε _m };

Update the user's cognitive data according to the user's response:

For each selected user j∈S _t we have:

n _t+1,j =n _t,j +1.

The block diagram of the implementation architecture of air-conditioning load aggregation participating in secondary frequency regulation is shown in Figure 2. When the system frequency is monitored below the threshold, the load aggregator receives the power regulation target issued by the grid dispatch center, and then based on the “risk aversion dobby” The "machine" learning framework selects the user to send adjustment instructions. If the user accepts the instruction, it corresponds to the air conditioning action, otherwise it does not act. The user's response will be fed back to the load aggregator in time to facilitate the formulation of the next user selection strategy.

Embodiment 2

An air-conditioning aggregation control system, comprising: a load operation monitoring module, a system frequency sensing module, a power grid dispatch center, a load control center module, an air-conditioning local control module, a communication module, and the like

The load operation monitoring module is used for real-time estimation of the actual available adjustment reserve after the N air-conditioning loads are aggregated.

Specifically, the heat transfer model of a single air conditioner is:

The discrete form of this formula is:

After the aggregated loads of N air conditioners, the maximum secondary frequency regulation reserve that can be used in real time is estimated as:

The system frequency sensing module is used to sense the system frequency change in real time. Specifically, when the disturbance occurs, the system frequency deviation is:

|Δf(t)|=|f(t)-f _N |

Among them, f _N is the rated frequency of the grid.

The power grid dispatching center, based on the system frequency sensing module sensing the system frequency change, formulates the secondary frequency modulation power adjustment target D _tar,t .

Specifically, for the system frequency deviation |Δf(t)|, the secondary frequency modulation power adjustment target D _tar,t is calculated according to the secondary frequency modulation coefficient and the secondary frequency modulation reserve in the system:

where max(P _AC ) is the maximum adjustable reserve, and |Δf _mSFR | is the maximum frequency deviation that the system can withstand during secondary frequency modulation.

Therefore, the target D _{tar,t of the load side participating in the secondary frequency regulation power regulation at time t} should take the smaller value between the power value that needs to be regulated to restore the system frequency and the actual available regulation reserve, namely:

D _tar,t =min(k _S |Δf(t)|,P _AC (t))

The load control center module is used to formulate a strategy selected by the user after receiving the secondary frequency regulation power adjustment target value assigned by the power grid dispatch center, send instructions to the user, and collect the user's feedback on the instruction at the same time.

Specifically, according to the behavior model that considers the uncertainty of the user's response to the secondary frequency regulation load regulation command, the user is strategically selected to send regulation commands based on the framework of "risk avoidance dobby", so as to reduce the actual power regulation amount D _{agg as much as possible, t} and secondary frequency modulation power adjust the target D _tar,t deviation target to achieve a more reliable secondary frequency modulation effect.

The load control center needs to be selected among N users

Each user sends an adjustment command. If the user accepts the adjustment command, the air conditioner will be temporarily turned off/on, but not all commands will be responded to at the same time. Even if the demand response contract is signed or the user is invited in advance, there will be Due to environmental or personal factors, some users will not respond to the adjustment command as expected but choose to withdraw, so that the actual power adjustment amount D _agg,t cannot match the secondary frequency modulation power adjustment target D _tar,t , and the more users who choose to withdraw, The greater the deviation between D _agg,t and D _tar,t . Therefore, a behavioral model considering the uncertainty of the user's response to the secondary FM load regulation command is established:

The response behavior of user i can be regarded as a random variable obeying the Bernoulli distribution, that is, the user has a probability of p _i to agree to accept the adjustment command, corresponding to the air-conditioning load action; there is a probability of (1- _pi ) to opt out of the response, corresponding to Air conditioner doesn't work.

The expected value is:

E(X _i )= _pi

The variance is:

σ _i ² = _pi (1- _pi )

Based on the user's response behavior model, in order to make the aggregated load as a reliable resource to achieve accurate secondary frequency regulation as much as possible, the goal of the load aggregator is to reduce the actual power regulation amount D _agg,t as much as possible compared to the secondary frequency The deviation of the FM power regulation target D _tar,t , so the load aggregation control target is described as:

min E(D _agg,t -D _tar,t ) ²

In order to achieve the above goals, the user's historical response feedback is firstly learned online to form the existing cognition of the user, including the probability estimation of the user's acceptance of adjustment instructions.

And the historical number of sent instructions n _t,i ; Combined with the above knowledge, before selecting users, the user is prioritized based on the framework of "risk avoidance dobby". Under this framework, on the one hand, the aggregator The existing knowledge of users should be fully utilized, and users with higher historical response power expectations and smaller variance should be selected. On the other hand, it is also necessary to explore the behavior of more users to form a broader understanding of different users. The trade-off with exploration can calculate the sorting index value v _t,i of each user; then select m users to send adjustment instructions according to the index value in descending order, until

like

Then let m=n, P ₁ , . . . , P _n respectively represent the air-conditioning power of the user, and n is the total number of users; after the command is sent, the user's cognition is updated online in real time according to the user's response result, as the next user The basis for selection, so a closed-loop process is formed.

variance estimate

And the number of times n _t,i of historically sent adjustment instructions;

The third step, at any time t∈[T]:

Calculate the sort index value for each user:

Sort users in descending order of the size of the index value v _t,i :

Select m users in sequence until:

like

Then let m=n;

Update the user's cognitive data according to the user's response:

For each selected user j∈S _t we have:

n _t+1,j =n _t,j +1.

The air conditioner local control module is used to execute the air conditioner adjustment command. Specifically, if the user accepts the adjustment instruction sent from the load control center, the corresponding switching operation is performed on the air conditioner.

The communication module is used to realize the information exchange between the grid dispatching center and the load control center, the load control center and the local control module of the air conditioner. Specifically, when the system frequency is monitored to be lower than the threshold, the load control center receives the power adjustment target issued by the power grid dispatch center; then the load control center selects the appropriate user through calculation to send the adjustment instruction; if the user accepts the instruction, the air conditioner local The control module performs the corresponding disconnection operation on the air conditioner, otherwise, if the user chooses to quit the response, the corresponding air conditioner local control module will not act; the user's response will be timely fed back to the load control center through the communication module, which is convenient for the user to select the strategy next time. formulate.

Embodiment 3

A control device includes a processor, and the processor is used to execute a program to implement the method of the first embodiment or the system flow of the second embodiment, thereby improving the reliability of load aggregation, realizing a more ideal frequency modulation effect, and avoiding blindness of aggregation quotients. Send adjustment instructions to users in a timely manner, reduce the number of users who send instructions and the number of users who opt out of responding, thereby reducing the cost of load aggregators and the fatigue of users due to frequent acceptance of adjustment instructions.

Embodiment 4:

The simulation is based on the improved IEEE RTS 24 node system. The system diagram is shown in Figure 3. The system reference power is 100MW, and the simulation environment is MATLAB PSAT (V.2.1.11). The processing methods of the first and second embodiments are verified. :

Assuming that there are 50,000 users in the system who have signed the frequency regulation agreement (one air conditioner per household, the average power of the air conditioner is 2.5kW), which are distributed at nodes #15, #19, and #20 respectively. When node #2 generates about 74MW of power generation If there is a shortage, the system frequency drops below the threshold of 59.97Hz, and a frequency modulation starts. When the system frequency returns to the new steady state, the frequency is about 59.9514Hz at this time, as shown in Figure 4.

In order to further restore the frequency to the normal level (60±0.04Hz), the air conditioner is required to participate in the secondary frequency modulation, and the reduction target of the secondary frequency modulation power is calculated to be 28.09MW at this time.

Figure 5 shows the comparison of the second frequency modulation results of the processing methods of Embodiments 1 and 2 with random selection and offline optimization methods (assuming that all user response behaviors are known in the ideal case), and the results show that the results are implemented in different frequency modulation events. The processing methods of Example 1 and Example 2 are obviously better than the random selection method in the effect of secondary frequency modulation.

Figures 6 and 7 show that in the process of continuously learning the user response behavior in the processing methods of the first and second embodiments, the load aggregation performance gradually approaches the offline optimization method, and the aggregation power gradually approaches the target value while the random However, there is always a large deviation between the selection method and the target adjustment amount;

At the same time, Figures 8 and 9 show that the random selection method sends more adjustment instructions to more users than the proposed method in all FM events, and nearly half of the users opt out. In practice, choosing more users means spending more incentive costs, and every time a user receives an instruction, it will increase their fatigue level, so fewer users to choose and users to quit can not only ensure the economy of load aggregation control It also improves user-friendliness, which is beneficial to both load aggregators and users.

Those skilled in the art can easily understand that the above description is only an embodiment of the present invention, and is not intended to limit the present invention, but any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention, etc. should be included within the protection scope of the present invention.

Claims

A kind of air conditioning polymerization control method, is characterized in that, comprises the following steps:

According to the load aggregation control objective, based on the historical feedback of the user's response to the adjustment instruction, the probability of accepting the adjustment instruction in the historical feedback is estimated
Higher, response behavior variance
It is relatively small and the number of historical times n t,i of sent instructions is less as a reference, set the user's sorting index value v t,i , and select m users in descending order of the sorting index value v t,i to send adjustment instructions ,until
like
Then let m=n, P 1 , . . . , P n respectively represent the air-conditioning power of users, and n is the total number of users.
The control method according to claim 1, wherein the sorting index value of the user is
The control method according to claim 1, wherein the load aggregation control objective is to minimize the deviation of the actual power adjustment amount D agg,t compared to the secondary frequency modulation target D tar,t .
The control method according to claim 3, characterized in that, at time t, the secondary frequency modulation power adjustment target D tar,t is between the adjusted power value required to restore the system frequency air conditioner and the actual available adjustment reserve the smaller value of .
The control method according to claim 4, wherein the power value to be adjusted to restore the system frequency air conditioner is the product of the secondary frequency modulation coefficient k S and the system frequency deviation |Δf(t)|.
The control method according to claim 4, wherein the actual available adjustment reserve in real time is
η represents the cooling coefficient of the air conditioner, s i (t) represents the switching state of the ith air conditioner at time t, and Qi represents the thermal power of the ith air conditioner.
The control method according to claim 1, characterized in that, after sending adjustment instructions to the users selected according to the index value sorting according to the load aggregation control target, the action information of the users in response to the adjustment instructions is collected and used for the next user selection. The sort index value is optimized.
An air-conditioning aggregation control system, characterized in that it comprises: a load operation monitoring module, a system frequency sensing module, a power grid dispatch center, a load control center module, an air-conditioning local control module and a communication module;

The load operation monitoring module is used for real-time estimation of the actual available adjustment reserve after the N air-conditioning loads are aggregated;

The system frequency sensing module is used to sense the system frequency change in real time;

The power grid dispatch center, based on the system frequency sensing module sensing the system frequency change and the actual available regulation reserve of the load operation monitoring module, formulates the secondary frequency regulation power regulation target;

The load control center module is used to formulate the strategy selected by the user after receiving the secondary frequency modulation power modulation target value distributed from the power grid dispatching center, and send the adjustment instruction to the user, and simultaneously collect the user's feedback on the instruction;

The local control module of the air conditioner performs corresponding switching operations on the air conditioner according to the user's response to the adjustment command of the load control center;

The communication module is used to realize the information exchange between the grid dispatching center and the load control center, the load control center and the local control module of the air conditioner.
A control device includes a processor, and the processor is used to execute a program to implement the method of any one of the claims 1-7 or the system flow of the claim 8.