Air conditioner temperature control load cluster interaction method and device

Abstract

The application provides an air conditioner temperature control load cluster interaction method and device. The method comprises the following steps: first, obtaining training samples and corresponding labeling information, and iteratively training a deep belief network based on the training samples and the labeling information to obtain a target network model; second, obtaining a target power consumption adjustment amount; and finally, processing the target power consumption adjustment amount, a target user's corresponding indoor average temperature and an outdoor average temperature through the target network model to obtain a target temperature update value, so as to set the temperature of the air conditioner corresponding to the target user. The above scheme uses a large amount of data samples as a driving force, combines the deep belief network to train the samples, and reasonably sets the temperature of the air conditioner corresponding to the target user, thereby realizing the interaction of the offline training and online interaction of the air conditioner temperature control load cluster, meeting the accuracy scheduling requirement of the air conditioner temperature control load cluster interaction, and being high in precision and strong in timeliness.

Description

Technical Field

[0001] This application relates to the field of temperature-controlled load cluster control technology, specifically to an interactive method and apparatus for air conditioning temperature-controlled load clusters. Background Technology

[0002] As a type of thermostatically controlled load (TCL), air conditioning load can convert electrical energy into heat energy stored in the room. This energy conversion and storage characteristic makes air conditioning the load with the greatest demand response potential. By reasonably controlling the air conditioning load, the peak load can be reduced and the supply and demand contradiction can be alleviated at a lower cost without affecting or with minimal impact on user comfort.

[0003] However, for a single air conditioner, its load characteristic curve varies depending on the type and parameters, and its power is relatively small. In contrast, aggregated air conditioning loads have the advantages of a considerable number, flexible scheduling methods, and huge potential for participating in system scheduling. This aggregation mode is generally implemented through a load aggregator mechanism. Currently, scholars at home and abroad have conducted a lot of research on the adjustability potential of aggregated air conditioning loads, but there is still considerable room for improvement. For example, existing technologies only consider the impact of a single factor on the air conditioner's state, ignoring the randomness of other interactive factors; or they use mathematical modeling to optimize the demand response of the temperature-controlled load side, but to some extent ignore the physical characteristics of the temperature-controlled load; or they optimize the scheduling plan and control strategy of the air conditioning load based on the air conditioning load output and bid of the load aggregator in each time period, but the uncertainty of the initial state of the air conditioner and the timeliness of the control need further optimization; or they regulate the temperature-controlled load cluster through intra-group control strategies, but its timeliness is difficult to guarantee.

[0004] Therefore, due to the lack of analysis and control methods, existing technologies cannot meet the requirements for accurate scheduling of air conditioning temperature control load cluster interaction. Summary of the Invention

[0005] This application provides a method and apparatus for interaction of air conditioning temperature control load clusters, which meets the accuracy scheduling requirements for interaction of air conditioning temperature control load clusters. The technical solution is as follows.

[0006] On the one hand, a method for interaction of air conditioning temperature control load clusters is provided, the method comprising:

[0007] Acquire training samples and corresponding annotation information; the training samples include simulated indoor temperature, simulated outdoor temperature, and power adjustment within a specified time period; the annotation information is the simulated temperature update value obtained under the conditions corresponding to the training samples.

[0008] The deep belief network is iteratively trained based on the training samples and the annotation information to obtain the target network model.

[0009] Obtain the target power consumption adjustment amount;

[0010] The target power consumption adjustment, the indoor average temperature and outdoor average temperature corresponding to the target user are processed through the target network model to obtain the target temperature update value, so as to set the temperature of the air conditioner corresponding to the target user.

[0011] On another front, an air conditioning temperature control load cluster interaction device is provided, the device comprising:

[0012] The sample and annotation acquisition module is used to acquire training samples and corresponding annotation information; the training samples include simulated indoor temperature, simulated outdoor temperature, and power adjustment amount within a specified time period; the annotation information is the simulated temperature update value obtained under the conditions corresponding to the training samples.

[0013] The target network model acquisition module is used to iteratively train the deep belief network based on the training samples and the annotation information to obtain the target network model.

[0014] The adjustment amount acquisition module is used to acquire the target power adjustment amount;

[0015] The target temperature update value acquisition module is used to process the target power consumption adjustment, the indoor average temperature corresponding to the target user, and the outdoor average temperature through the target network model to obtain the target temperature update value, so as to set the temperature of the air conditioner corresponding to the target user.

[0016] In one possible implementation, the sample and annotation acquisition module includes:

[0017] The temperature selection unit is used to sequentially select the simulated indoor temperature and the simulated outdoor temperature according to a specified step size;

[0018] A temperature value acquisition unit is set to set various power adjustment amounts for each pair of simulated indoor temperature and simulated outdoor temperature within a specified time period, and to perform temperature regulation simulation processing on each test air conditioner to obtain the set temperature value of each test air conditioner.

[0019] The update value determination unit is used to determine the simulated temperature update value based on the set temperature value of each test air conditioner.

[0020] In one possible implementation, the set temperature value acquisition unit is further configured to:

[0021] For each power adjustment, under the conditions of simulated indoor temperature and simulated outdoor temperature, the power adjustment is set, and temperature regulation simulation processing is performed on each test air conditioner according to the operating cooling capacity, equivalent resistance and equivalent capacitance of each test air conditioner to obtain the set temperature value of each test air conditioner.

[0022] In one possible implementation, the update value determination unit is further configured to:

[0023] The set temperature values ​​of each of the test air conditioners are determined as the simulated temperature update values.

[0024] In one possible implementation, the target network model acquisition module includes:

[0025] The normalization processing unit is used to normalize the training samples and the corresponding annotation information to obtain normalized training samples and normalized annotation information.

[0026] The iterative training unit is used to iteratively train the deep belief network based on the normalized training samples and normalized annotation information to obtain the target network model.

[0027] In one possible implementation, the iterative training unit is further configured to:

[0028] The normalized training samples are processed by the deep belief network to obtain prediction results;

[0029] Calculate the root mean square error, mean absolute error, and coefficient of determination between the prediction result and the normalized annotation information, respectively.

[0030] When at least one of the root mean square error, mean absolute error, and coefficient of determination does not meet the specified conditions, the loss function value is obtained based on the prediction result and the annotation information, and the parameters of the deep belief network are updated based on the loss function value, and the training process is repeated iteratively.

[0031] When the root mean square error, mean absolute error, and coefficient of determination meet the specified conditions, the updated deep belief network is determined as the target network model.

[0032] In one possible implementation, the device is further configured to:

[0033] Obtain the estimated dispatched power volume sent by the load aggregator;

[0034] The ratio of the expected scheduled power consumption to the specified duration is determined as the target power consumption adjustment amount.

[0035] In another aspect, a computer device is provided, the computer device including a processor and a memory, the memory storing at least one instruction, the at least one instruction being loaded and executed by the processor to implement an air conditioning temperature control load cluster interaction method as described above.

[0036] In another aspect, a computer-readable storage medium is provided, wherein at least one instruction is stored therein, the at least one instruction being loaded and executed by a processor to implement an air conditioning temperature control load cluster interaction method as described above.

[0037] The technical solution provided in this application may include the following beneficial effects:

[0038] First, training samples and corresponding annotation information are acquired. Based on these samples and annotation information, a deep belief network is iteratively trained to obtain the target network model. Next, the target power consumption adjustment amount is acquired. Finally, the target network model is used to process the target power consumption adjustment amount, the average indoor temperature of the target user, and the average outdoor temperature to obtain the target temperature update value, which is then used to set the temperature of the air conditioner corresponding to the target user. This approach uses a large number of data samples as the driving force, combined with a deep belief network to train the samples, to set reasonable temperatures for the air conditioners corresponding to the target users. This achieves offline training and online interaction for the air conditioning temperature control load cluster, meeting the accuracy scheduling requirements for the interaction of the air conditioning temperature control load cluster with high precision and timeliness. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0040] Figure 1 This is a schematic diagram of the structure of an air conditioning temperature control load cluster interaction system according to an exemplary embodiment.

[0041] Figure 2 This is a flowchart illustrating an interactive method for air conditioning temperature control load clusters according to an exemplary embodiment.

[0042] Figure 3 This is a flowchart illustrating an interactive method for air conditioning temperature control load clusters according to an exemplary embodiment.

[0043] Figure 4 A schematic diagram of the structure of a deep belief network involved in an embodiment of this application is shown.

[0044] Figure 5 This is a comparison curve of test results for an air conditioning temperature control load cluster under different training models, as illustrated in an exemplary embodiment.

[0045] Figure 6 An interactive flowchart illustrating the adjustable potential of an air conditioning temperature control load cluster is shown in an exemplary embodiment of this application.

[0046] Figure 7 The diagram illustrates a schematic representation of the changes in photovoltaic output, base load, total load, and outdoor temperature over time, as shown in an exemplary embodiment of this application.

[0047] Figure 8 A schematic diagram of the photovoltaic backfeed curves of the system operation before and after the addition of demand response is shown in an exemplary embodiment of this application.

[0048] Figure 9 A schematic diagram of indoor and outdoor temperature changes before and after the addition of demand response is shown in an exemplary embodiment of this application.

[0049] Figure 10 The diagram shows a structural schematic of an air conditioning temperature control load cluster interaction device according to an exemplary embodiment of this application.

[0050] Figure 11 A structural block diagram of a computer device illustrated in an exemplary embodiment of this application is shown. Detailed Implementation

[0051] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0052] It should be understood that in the description of the embodiments of this application, the term "correspondence" may indicate that there is a direct or indirect correspondence between the two, or that there is an association between the two, or that there is a relationship of instruction and being instructed, configuration and being configured, etc.

[0053] Figure 1 This is a schematic diagram illustrating the structure of an air conditioning temperature control load cluster interactive system according to an exemplary embodiment. The system includes a server 110 and an air conditioning temperature control load cluster 120. The air conditioning temperature control load cluster 120 includes multiple test air conditioners.

[0054] Optionally, the air conditioning temperature control load cluster 120 can communicate with the server 110 through a transmission network (such as a wireless communication network). The air conditioning temperature control load cluster 120 can upload data (such as the set temperature) to the server 110 through the wireless communication network so that the server 110 can process it.

[0055] Optionally, the server 110 calculates new set temperatures for the air conditioning cluster under different outdoor average temperatures, different indoor set temperatures, and different power adjustment combinations, and finally obtains each set of simulation data corresponding to the air conditioning temperature control load cluster 120. Then, based on each set of simulation data, training samples and corresponding annotation information are generated and trained to obtain the target network model, thereby enabling the setting of temperature for any test air conditioner in the air conditioning temperature control load cluster 120.

[0056] Optionally, the server 110 can also establish a communication connection with the air conditioning temperature control load cluster 120 through a wireless communication network and send corresponding algorithm information to each test air conditioner. For example, the server 110 can calculate the target temperature update value corresponding to the target test air conditioner according to the trained target network model, and send it to the air conditioning temperature control load cluster 120 through the wireless communication network to set the temperature of the target test air conditioner.

[0057] Optionally, the aforementioned server 110 can be a server cluster or a distributed system composed of multiple physical servers, or it can be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms and other technology computing services.

[0058] Optionally, the system may also include a management device for managing the system (such as managing the connection status between each module and the server), and the management device is connected to the server via a communication network. Optionally, the communication network may be a wired network or a wireless network.

[0059] Optionally, the aforementioned wireless or wired networks use standard communication technologies and / or protocols. The network is typically the Internet, but can also be any other network, including but not limited to any combination of local area networks (LANs), metropolitan area networks (MANs), wide area networks (WANs), mobile, wired or wireless networks, private networks, or analog private networks. In some embodiments, technologies and / or formats including Hypertext Markup Language (HTML), Extensible Markup Language (XML), etc., are used to represent data exchanged over the network. Furthermore, conventional encryption technologies such as Secure Sockets Layer (SSL), Transport Layer Security (TLS), Analog Private Networks (APN), and Internet Protocol Security (IPS) can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can be used to replace or supplement the aforementioned data communication technologies.

[0060] Figure 2 This is a flowchart illustrating an interactive method for air conditioning temperature control load clusters according to an exemplary embodiment. The method is executed by a computer device, which may be, for example... Figure 1 Server 110 is shown in the image. Figure 2 As shown, the air conditioning temperature control load cluster interaction method may include the following steps:

[0061] S201. Obtain training samples and corresponding annotation information; the training samples include simulated indoor temperature, simulated outdoor temperature, and power adjustment within a specified time period; the annotation information is the simulated temperature update value obtained under the conditions corresponding to the training samples.

[0062] In one possible implementation, when performing interaction of the air conditioning temperature control load cluster, based on the simulation algorithm of the air conditioning temperature control load cluster, the set temperature (i.e., the simulated outdoor temperature) of each test air conditioner in the air conditioning temperature control load cluster is calculated under different outdoor average temperatures (i.e., the simulated outdoor temperature), different indoor set temperatures (i.e., the simulated indoor temperature), and different combinations of power adjustment amounts. The simulated indoor temperature, the simulated outdoor temperature, and the power adjustment amount within a specified time period are used as input features of the temperature interaction sample, and the simulated temperature update value is used as the output feature of the temperature interaction sample to construct training samples for training.

[0063] Furthermore, since the event interval in a real power grid is usually 15 minutes, the specified duration can be 15 minutes.

[0064] S202. Based on the training samples and the annotation information, the deep belief network is iteratively trained to obtain the target network model.

[0065] In one possible implementation, a deep belief network is built based on the aforementioned input features, namely, simulated indoor temperature, simulated outdoor temperature, and power adjustment within a specified time period. The deep belief network is then trained offline based on training samples and labeled information. When the training conditions meet preset requirements, the model training ends and an optimal deep belief network model is generated. This optimal deep belief network model is the target network model to be built. By inputting the target user's demand response potential interaction data (i.e., indoor temperature, outdoor temperature, and power adjustment) into the target network model, the corresponding interaction results at the set temperature can be obtained.

[0066] S203. Obtain the target power adjustment amount.

[0067] In one possible implementation, the target power adjustment amount for the target user is determined based on the total power consumption that the load aggregator needs to schedule and the specified duration mentioned above.

[0068] S204. Using the target network model, process the target power consumption adjustment, the indoor average temperature corresponding to the target user, and the outdoor average temperature to obtain the target temperature update value, so as to set the temperature of the air conditioner corresponding to the target user.

[0069] In one possible implementation, when interacting with the demand response potential of target users using a target network model, a real-time output value, namely the target temperature update value, is obtained based on the real-time input of the target power adjustment amount, the indoor average temperature corresponding to the target user, and the outdoor average temperature. The target temperature update value is then used to set the temperature of the air conditioner corresponding to the target user, thereby realizing online interaction of the adjustable potential of the air conditioning temperature control load cluster.

[0070] In summary, the approach first acquires training samples and corresponding annotation information, then iteratively trains a deep belief network based on these samples and annotations to obtain the target network model. Next, it acquires the target power consumption adjustment amount. Finally, using the target network model, it processes the target power consumption adjustment amount, the average indoor temperature for the target user, and the average outdoor temperature to obtain the target temperature update value, which is then used to set the temperature of the air conditioner corresponding to the target user. This scheme uses a large amount of data samples as the driving force, combined with a deep belief network to train the samples, to set reasonable temperatures for the air conditioners corresponding to the target users. This achieves offline training and online interaction for the air conditioning temperature control load cluster, meeting the accuracy scheduling requirements for the interaction of the air conditioning temperature control load cluster with high precision and timeliness.

[0071] Figure 3 This is a flowchart illustrating an interactive method for air conditioning temperature control load clusters according to an exemplary embodiment. The method is executed by a computer device, which may be, for example... Figure 1 Server 110 is shown in the image. Figure 3 As shown, the air conditioning temperature control load cluster interaction method may include the following steps:

[0072] S301. Select the simulated indoor temperature and simulated outdoor temperature in sequence according to the specified step size.

[0073] Furthermore, when conducting interaction of air conditioning temperature control load clusters, a simulation algorithm for air conditioning temperature control load clusters is first constructed based on a first-order equivalent thermal parameter model. Considering the variation of outdoor average temperature and indoor average temperature within a certain range, simulated indoor temperature and simulated outdoor temperature are selected sequentially. The simulated outdoor temperature is considered to be 20℃ to 38℃, increasing in increments of 0.5℃; the simulated indoor average temperature is considered to be 20℃ to 28℃, increasing in increments of 1℃.

[0074] S302. For each pair of simulated indoor and simulated outdoor temperatures, set various power adjustment amounts within a specified time period, and perform temperature regulation simulation processing on each test air conditioner to obtain the set temperature value of each test air conditioner.

[0075] In one possible implementation, for each power adjustment, under the conditions of the simulated indoor temperature and the simulated outdoor temperature, the power adjustment is set, and temperature regulation simulation processing is performed on each test air conditioner according to the operating cooling capacity, equivalent resistance, and equivalent capacitance of each test air conditioner to obtain the set temperature value of each test air conditioner.

[0076] Furthermore, considering that the power adjustment of each test air conditioner in the air conditioning temperature control load cluster is from -2kW to 2kW, increasing in increments of 0.1kW, it is stipulated here that the power adjustment is negative when adjusting the temperature downward and positive when adjusting the temperature upward; and considering that the event interval in the actual power grid is 15min, the specified duration can be set to 15min.

[0077] Furthermore, by utilizing the formula derived from the first-order equivalent thermal parameter model, a simulation algorithm for air conditioning temperature-controlled load clusters is constructed. By using this simulation algorithm to adjust the power of the air conditioning temperature-controlled load clusters across the entire range, the set temperature values ​​corresponding to the air conditioning temperature-controlled load clusters can be obtained in batches.

[0078] The formula for calculating the set temperature value of a test air conditioner is:

[0079]

[0080] The formula for calculating the set temperature value of an air conditioning temperature control load cluster is as follows:

[0081] T s =min(T) ss );

[0082] Among them, T out The simulated outdoor average temperature is represented by Q, the operating cooling capacity of the tested air conditioner is represented by R and C, respectively, and T represents the equivalent thermal resistance and equivalent heat capacity. inThis represents the simulated average indoor temperature, for example, 24℃, and Δt represents the specified duration, which is fixed at 15 minutes.

[0083] S303. Determine the simulated temperature update value based on the set temperature value of each test air conditioner.

[0084] In one possible implementation, the set temperature value of each test air conditioner is determined as the simulated temperature update value.

[0085] S304. Iteratively train the deep belief network based on training samples and labeled information to obtain the target network model.

[0086] In one possible implementation, the training samples and their corresponding annotation information are normalized to obtain normalized training samples and normalized annotation information.

[0087] Based on the normalized training samples and normalized annotation information, the deep belief network is iteratively trained to obtain the target network model.

[0088] In one possible implementation, the normalized training samples are processed by the deep belief network to obtain the prediction results;

[0089] Calculate the root mean square error, mean absolute error, and coefficient of determination between the prediction result and the normalized annotation information, respectively.

[0090] When at least one of the root mean square error, mean absolute error, and coefficient of determination does not meet the specified conditions, the loss function value is obtained based on the prediction result and the annotation information, and the parameters of the deep belief network are updated based on the loss function value, and the training process is repeated iteratively.

[0091] When the root mean square error, mean absolute error, and coefficient of determination meet the specified conditions, the updated deep belief network is determined as the target network model.

[0092] Furthermore, the simulated temperature update values ​​are preprocessed. The simulated outdoor average temperature sequence, the simulated indoor average temperature sequence, and the power adjustment within 15 minutes are selected as the input features of the interactive samples, and the output features are the simulated temperature update values. A sample set is constructed and normalized. The normalization process is as follows: all the initial training samples (including the above training samples and corresponding annotation information) are normalized to the interval between 0 and 1 to obtain normalized training samples and normalized annotation information, which is convenient for the subsequent training of the deep belief network.

[0093] Furthermore, a deep belief network is constructed based on the aforementioned input and output features. The training sample data is divided into a training set and a test set in a 4:1 ratio. The error rate and the coefficient of determination are selected as interaction indicators to construct an interaction model of the maximum adjustable potential of the air conditioning temperature control load cluster (this interaction model is also the target network model). The structure of this interaction model consists of an input layer INPUT, four restricted Boltzmann machine layers (1RBM, RBM2, RBM3, RBM4), and an output layer OUTPUT.

[0094] Furthermore, the constructed samples are trained offline using a deep belief network. When the training set error rate and coefficient of determination meet preset requirements, model training ends, and the optimal model weight parameters are saved. When training the deep belief network model, RMSE is defined as the root mean square error between the predicted and actual samples.

[0095]

[0096] Define MAE as the mean absolute error between the predicted sample and the actual sample:

[0097]

[0098] Define R² as the coefficient of determination for the model:

[0099]

[0100] Where m represents the total number of test set samples in the training samples, and P i Represents the actual sample value. This represents the sample value predicted by the deep belief network. This is the average of the actual sample values.

[0101] RMSE and MAE are used as the standard for predicting the predictive ability of the interactive model, and the coefficient of determination R2 is used as the evaluation standard for the goodness of model fit. When the two error rates and the coefficient of determination meet the threshold requirements (i.e., the conditions specified above), the model training ends, and the optimal model parameters (updated deep belief network parameters) can be obtained. The optimal deep belief model is then determined as the target network model. If the conditions are not met, the loss function value is obtained, and the parameters of the deep belief network are updated based on the loss function value. The training process is repeated iteratively until the preset threshold standard is met.

[0102] S305. Obtain the target power adjustment amount.

[0103] In one possible implementation, the expected dispatched power volume sent by the load aggregator is obtained;

[0104] The ratio of the expected dispatched power to the specified duration is determined as the target power adjustment amount.

[0105] Furthermore, based on the equal area rule, the total electricity consumption that the load aggregator wants to schedule is converted into the power adjustment amount within 15 minutes in real time. According to the target network model obtained above, the interactive data of the target user demand response potential is input into the already trained deep belief network model (i.e., the target network model) to obtain the corresponding interactive results of the set temperature. The network is used to conduct online interaction with the target user with adjustable potential. The specific calculation formula of the equal area rule is: W = P·Δt, where W is the total electricity consumption that the load aggregator needs to schedule based on the air conditioning temperature control load cluster response potential assessment model, Δt is 15 minutes, and P can be obtained as the power adjustment amount within 15 minutes, that is, the target power adjustment amount.

[0106] S306. Using the target network model, process the target power consumption adjustment, the indoor average temperature corresponding to the target user, and the outdoor average temperature to obtain the target temperature update value, so as to set the temperature of the air conditioner corresponding to the target user.

[0107] In one possible implementation, by using an optimized deep belief network (i.e., the target network model), a real-time output value, i.e., the target temperature update value, is obtained from the real-time input outdoor average temperature, indoor average temperature, and 15-minute power adjustment sample, so as to realize the online interaction of the adjustable potential of the air conditioning temperature control load cluster.

[0108] In summary, the approach first acquires training samples and corresponding annotation information, then iteratively trains a deep belief network based on these samples and annotations to obtain the target network model. Next, it acquires the target power consumption adjustment amount. Finally, using the target network model, it processes the target power consumption adjustment amount, the average indoor temperature for the target user, and the average outdoor temperature to obtain the target temperature update value, which is then used to set the temperature of the air conditioner corresponding to the target user. This scheme uses a large amount of data samples as the driving force, combined with a deep belief network to train the samples, to set reasonable temperatures for the air conditioners corresponding to the target users. This achieves offline training and online interaction for the air conditioning temperature control load cluster, meeting the accuracy scheduling requirements for the interaction of the air conditioning temperature control load cluster with high precision and timeliness.

[0109] The following simple example illustrates the content disclosed in the above embodiments. Taking a cluster of 1500 air conditioner temperature-controlled loads as an example, the outdoor temperature is set from 20℃ to 38℃, increasing in increments of 0.5℃; the power adjustment of each air conditioner is from -2kW to 2kW, increasing in increments of 0.1kW; the indoor set temperature is set from 20℃ to 28℃, increasing in increments of 1℃; using a first-order equivalent thermal parameter model, the corresponding set temperature of the air conditioner load cluster is calculated. Similarly, the simulated temperature update values ​​of the air conditioner cluster under different simulated average outdoor temperatures, different simulated indoor set temperatures, and different combinations of power adjustments can be calculated, ultimately yielding 13653 sets of simulation data.

[0110] The simulation data is preprocessed by selecting the simulated outdoor average temperature sequence, the simulated indoor average temperature sequence, and the power adjustment within 15 minutes as the input features of the interactive samples. The output feature is the new set temperature value of the air conditioner. A sample set is constructed and normalized, that is, all the obtained initial sample data (including training samples and corresponding annotation information) are normalized to the range of 0 to 1, which facilitates the subsequent training of the deep belief network and ensures the integrity and accuracy of the sample data.

[0111] The preprocessed training sample data was divided into a training set and a test set in a 4:1 ratio, with 11,000 training samples and 2,653 validation samples. The constructed sample data was then placed into... Figure 4 The deep belief network shown is trained offline. Figure 4 The deep belief network structure consists of one input layer, four hidden layers, and one output layer. The input layer takes the average outdoor temperature T as input. out The set temperature value T of the air conditioning temperature control load cluster ss And the power adjustment amount P, the output layer outputs the set temperature value T of the temperature control load cluster of a test air conditioner. s .

[0112] Next, to measure the effectiveness of the deep learning algorithm on the potential interaction of air conditioning temperature control load clusters, RMSE is defined as the root mean square error between predicted and actual samples, MAE is defined as the mean absolute error between predicted and actual samples, and R² is defined as the model's coefficient of determination. These are calculated separately. When both error rates and the coefficient of determination meet the threshold requirements, model training ends, and the optimal deep belief network parameters are obtained. This optimal deep belief model is then determined as the target network model. Finally, for the real-time input of simulated outdoor average temperature, simulated indoor average temperature, and 15-minute power adjustment samples, the set temperature error rate and coefficient of determination for real-time interaction are shown in Table 1, which meet the preset threshold requirements.

[0113] Table 1

[0114]

[0115] As shown in Table 1, the root mean square error and mean absolute error of the real-time interactive set temperature values ​​trained under the deep belief network model both reach the level of 10^-3, indicating that the prediction results of the deep belief network model are quite accurate. Meanwhile, the coefficient of determination R², which measures the goodness of fit of the interactive model, is also close to 1, indicating that the deep belief network model has a good fit.

[0116] Table 2 presents a comparison of test results for air conditioning temperature control load clusters under different training models:

[0117] Table 2

[0118]

[0119] Please refer to Figure 5 The graph shows a comparison of test results for air conditioning temperature control load clusters under different training models. Figure 5 The comparison shows that the interaction performance of the Deep Belief Network (DBN) model is significantly higher than that of the Support K Nearest Neighbors (KNN), Decision Tree (DT), Multilayer Perceptron (MLP), and Vector Machine (SVM) models.

[0120] Please refer to Figure 6 The diagram illustrates an interactive flowchart illustrating the adjustable potential of an air conditioning temperature-controlled load cluster, based on... Figure 6 The training process yields an adjustable potential interaction model (i.e., the target network model). By inputting the current real-time outdoor temperature and indoor average temperature data, load aggregators can achieve online interaction of the adjustable potential of the corresponding air conditioning temperature control load cluster.

[0121] To further verify the positive impact of the adjustability potential of air conditioning temperature control load clusters based on data-driven and deep belief networks on photovoltaic backfeed phenomena and grid operation stability, Figure 7 Taking the actual operating data of a power distribution network during a certain period as an example, the actual sampling interval is 15 minutes, with a total of 96 time periods. The actual air temperature is used as the outside temperature. Curve 1 represents photovoltaic output, curve 2 represents base load, curve 3 represents total load, and curve 4 represents outdoor temperature. Figure 7 As shown, the highest outdoor temperature is 34℃. In this example, all 1500 air conditioning units in the cluster participate in demand response. Using the interaction model between the air conditioning cluster's potential assessment model and its response characteristics, the calculation of photovoltaic backfeed is defined as the difference between the total photovoltaic output and the total load. The photovoltaic backfeed curves before and after incorporating demand response, as well as the indoor and outdoor temperature change curves, are obtained as follows: Figure 8 , Figure 9 As shown.

[0122] like Figure 8 As shown, curve 1 represents the power curve of the air conditioning cluster without demand response interactive scheduling, which exhibits a clear period of photovoltaic backfeed. Curve 2 represents the power curve of a 1500-unit air conditioning load cluster after real-time demand response interaction based on the load aggregator's assessment of the air conditioning cluster's potential. From the photovoltaic backfeed situation before and after adding demand response interaction, it can be seen that the power curve of the air conditioning cluster without demand response interactive scheduling exhibits a clear period of photovoltaic backfeed. Figure 8 The curve for air conditioning load clusters above 0KW (1), while the power curve of the air conditioning load cluster after real-time interaction with the load aggregator's assessment of the potential of the air conditioning cluster, can completely absorb the period when photovoltaic backfeed occurs ( Figure 8 Curve 2 remains at or below 0kW, during which time the maximum photovoltaic capacity can reach 208kW, and its control time is rapid; Figure 9 As shown, the changes in indoor temperature before and after the addition of demand response interaction can be seen from the fact that the photovoltaic backfeed phenomenon can be eliminated after the controlled indoor temperature drops from 26℃ to 22℃. This proves that the demand response interaction model (i.e., the target network model) of the adjustable potential of the air conditioning temperature control load cluster described in this application can indeed effectively improve the power consumption of the air conditioning load cluster, solve the photovoltaic backfeed problem, and improve the stability of power system operation.

[0123] Figure 10 This is a structural block diagram illustrating an air conditioning temperature control load cluster interaction device according to an exemplary embodiment. The device includes:

[0124] The sample and annotation acquisition module 101 is used to acquire training samples and corresponding annotation information; the training samples include simulated indoor temperature, simulated outdoor temperature and power adjustment amount within a specified time period; the annotation information is the simulated temperature update value obtained under the conditions corresponding to the training samples.

[0125] The target network model acquisition module 102 is used to iteratively train the deep belief network based on the training samples and the annotation information to obtain the target network model.

[0126] The adjustment amount acquisition module 103 is used to acquire the target power consumption adjustment amount;

[0127] The target temperature update value acquisition module 104 is used to process the target power consumption adjustment amount, the indoor average temperature corresponding to the target user, and the outdoor average temperature through the target network model to obtain the target temperature update value, so as to set the temperature of the air conditioner corresponding to the target user.

[0128] In one possible implementation, the sample and annotation acquisition module 101 includes:

[0129] The temperature selection unit is used to sequentially select the simulated indoor temperature and the simulated outdoor temperature according to a specified step size;

[0130] A temperature value acquisition unit is set to set various power adjustment amounts for each pair of simulated indoor temperature and simulated outdoor temperature within a specified time period, and to perform temperature regulation simulation processing on each test air conditioner to obtain the set temperature value of each test air conditioner.

[0131] The update value determination unit is used to determine the simulated temperature update value based on the set temperature value of each test air conditioner.

[0132] In one possible implementation, the set temperature value acquisition unit is further configured to:

[0133] For each power adjustment, under the conditions of simulated indoor temperature and simulated outdoor temperature, the power adjustment is set, and temperature regulation simulation processing is performed on each test air conditioner according to the operating cooling capacity, equivalent resistance and equivalent capacitance of each test air conditioner to obtain the set temperature value of each test air conditioner.

[0134] In one possible implementation, the update value determination unit is further configured to:

[0135] The set temperature values ​​of each of the test air conditioners are determined as the simulated temperature update values.

[0136] In one possible implementation, the target network model acquisition module 102 includes:

[0137] The normalization processing unit is used to normalize the training samples and the corresponding annotation information to obtain normalized training samples and normalized annotation information.

[0138] The iterative training unit is used to iteratively train the deep belief network based on the normalized training samples and normalized annotation information to obtain the target network model.

[0139] In one possible implementation, the iterative training unit is further configured to:

[0140] The normalized training samples are processed by the deep belief network to obtain prediction results;

[0141] Calculate the root mean square error, mean absolute error, and coefficient of determination between the prediction result and the normalized annotation information, respectively.

[0142] When at least one of the root mean square error, mean absolute error, and coefficient of determination does not meet the specified conditions, the loss function value is obtained based on the prediction result and the annotation information, and the parameters of the deep belief network are updated based on the loss function value, and the training process is repeated iteratively.

[0143] When the root mean square error, mean absolute error, and coefficient of determination meet the specified conditions, the updated deep belief network is determined as the target network model.

[0144] In one possible implementation, the device is further configured to:

[0145] Obtain the estimated dispatched power volume sent by the load aggregator;

[0146] The ratio of the expected scheduled power consumption to the specified duration is determined as the target power consumption adjustment amount.

[0147] In summary, the approach first acquires training samples and corresponding annotation information, then iteratively trains a deep belief network based on these samples and annotations to obtain the target network model. Next, it acquires the target power consumption adjustment amount. Finally, using the target network model, it processes the target power consumption adjustment amount, the average indoor temperature for the target user, and the average outdoor temperature to obtain the target temperature update value, which is then used to set the temperature of the air conditioner corresponding to the target user. This scheme uses a large amount of data samples as the driving force, combined with a deep belief network to train the samples, to set reasonable temperatures for the air conditioners corresponding to the target users. This achieves offline training and online interaction for the air conditioning temperature control load cluster, meeting the accuracy scheduling requirements for the interaction of the air conditioning temperature control load cluster with high precision and timeliness.

[0148] Figure 11 This illustration shows a structural block diagram of a computer device according to an exemplary embodiment of this application. The computer device includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the aforementioned air conditioning temperature control load cluster interaction method.

[0149] The processor can be a central processing unit (CPU). It can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations thereof.

[0150] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the embodiments of this invention. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory, thereby implementing the methods described in the above embodiments.

[0151] The memory may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0152] One embodiment of this application also provides a computer storage medium for storing a computer program, which, when executed by a processor, implements the above-described method for interaction of air conditioning temperature control load clusters.

[0153] Those skilled in the art will understand that all or part of the processes in the above-described embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium can also include combinations of the above types of memory.

[0154] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for interaction among clusters of air conditioning temperature control loads, characterized in that, The method includes: Acquire training samples and corresponding annotation information; the training samples include simulated indoor temperature, simulated outdoor temperature, and power adjustment within a specified time period; the annotation information is the simulated temperature update value obtained under the conditions corresponding to the training samples. The deep belief network is iteratively trained based on the training samples and the annotation information to obtain the target network model. Obtain the target power consumption adjustment amount; The target power consumption adjustment, the indoor average temperature and outdoor average temperature corresponding to the target user are processed through the target network model to obtain the target temperature update value, so as to set the temperature of the air conditioner corresponding to the target user. The acquisition of training samples and corresponding annotation information includes: Select the simulated indoor temperature and simulated outdoor temperature in sequence according to the specified step size; For each pair of simulated indoor and simulated outdoor temperatures, various power adjustment amounts are set within a specified time period, and temperature regulation simulation processing is performed on each test air conditioner to obtain the set temperature value of each test air conditioner; the step of setting various power adjustment amounts within a specified time period and performing temperature regulation simulation processing on each test air conditioner to obtain the set temperature value of each test air conditioner includes: for each power adjustment amount, under the conditions of the simulated indoor and simulated outdoor temperatures, setting the power adjustment amount, and performing temperature regulation simulation processing on each test air conditioner according to the operating cooling capacity, equivalent resistance, and equivalent capacitance of each test air conditioner to obtain the set temperature value of each test air conditioner. The simulated temperature update value is determined based on the set temperature value of each test air conditioner.

2. The method according to claim 1, characterized in that, Based on the set temperature values ​​of each of the test air conditioners, the simulated temperature update value is determined, including: The set temperature values ​​of each of the test air conditioners are determined as the simulated temperature update values.

3. The method according to any one of claims 1 to 2, characterized in that, The iterative training of the deep belief network based on the training samples and the labeled information to obtain the target network model includes: The training samples and their corresponding annotation information are normalized to obtain normalized training samples and normalized annotation information. Based on the normalized training samples and normalized annotation information, the deep belief network is iteratively trained to obtain the target network model.

4. The method according to claim 3, characterized in that, The step of iteratively training the deep belief network based on the normalized training samples and normalized annotation information to obtain the target network model includes: The normalized training samples are processed by the deep belief network to obtain prediction results; Calculate the root mean square error, mean absolute error, and coefficient of determination between the prediction result and the normalized annotation information, respectively. When at least one of the root mean square error, mean absolute error, and coefficient of determination does not meet the specified conditions, the loss function value is obtained based on the prediction result and the annotation information, and the parameters of the deep belief network are updated based on the loss function value, and the training process is repeated iteratively. When the root mean square error, mean absolute error, and coefficient of determination meet the specified conditions, the updated deep belief network is determined as the target network model.

5. The method according to any one of claims 1 to 2, characterized in that, Before processing the target power consumption adjustment, the indoor average temperature corresponding to the target user, and the outdoor average temperature through the target network model to obtain the target temperature update value, the method further includes: Obtain the estimated dispatched power volume sent by the load aggregator; The ratio of the expected scheduled power consumption to the specified duration is determined as the target power consumption adjustment amount.

6. An air conditioning temperature control load cluster interaction device, characterized in that, The device includes: The sample and annotation acquisition module is used to acquire training samples and corresponding annotation information; the training samples include simulated indoor temperature, simulated outdoor temperature, and power adjustment amount within a specified time period; the annotation information is the simulated temperature update value obtained under the conditions corresponding to the training samples. The target network model acquisition module is used to iteratively train the deep belief network based on the training samples and the annotation information to obtain the target network model. The adjustment amount acquisition module is used to acquire the target power adjustment amount; The target temperature update value acquisition module is used to process the target power consumption adjustment amount, the indoor average temperature and outdoor average temperature corresponding to the target user through the target network model to obtain the target temperature update value, so as to set the temperature of the air conditioner corresponding to the target user. The sample and annotation acquisition module includes: The temperature selection unit is used to sequentially select the simulated indoor temperature and the simulated outdoor temperature according to a specified step size; A temperature value acquisition unit is set to set various power adjustment amounts for each pair of simulated indoor temperature and simulated outdoor temperature within a specified time period, and to perform temperature regulation simulation processing on each test air conditioner to obtain the set temperature value of each test air conditioner. The update value determination unit is used to determine the simulated temperature update value based on the set temperature value of each test air conditioner. The set temperature value acquisition unit is further configured to: For each power adjustment, under the conditions of simulated indoor temperature and simulated outdoor temperature, the power adjustment is set, and temperature regulation simulation processing is performed on each test air conditioner according to the operating cooling capacity, equivalent resistance and equivalent capacitance of each test air conditioner to obtain the set temperature value of each test air conditioner.

7. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing at least one instruction, which is loaded and executed by the processor to implement an air conditioning temperature control load cluster interaction method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The storage medium stores at least one instruction, which is loaded and executed by a processor to implement an air conditioning temperature control load cluster interaction method as described in any one of claims 1 to 5.

Images