Building energy consumption control method, device and equipment and storage medium

Abstract

The invention discloses a building energy consumption control method, device and equipment and a storage medium, and relates to the technical field of building management. The method comprises the following steps: acquiring energy consumption associated data of a current building, and determining predicted energy consumption of the current building by using a preset neural network according to the energy consumption associated data; for different types of electric equipment in the current building, generating an energy-saving control strategy by using a mixed integer linear programming model according to the predicted energy consumption corresponding to the electric equipment and a preset constraint condition; and controlling the electric equipment according to the energy-saving control strategy and historical energy-saving control strategy feedback data. According to the technical scheme of the embodiment of the invention, the prediction accuracy is improved, the performability of the control strategy is improved, and a full-link intelligent energy-saving system with high performability is constructed.

Description

Technical Field

[0001] This invention relates to the field of building management technology, and in particular to a building energy consumption control method, device, equipment and storage medium. Background Technology

[0002] With the acceleration of urbanization and the continuous expansion of building scale, the proportion of building energy consumption in global energy consumption continues to rise. Statistics show that commercial buildings and public facilities account for more than 40% of total social electricity consumption, with the energy consumption of equipment such as air conditioning, lighting, and elevators being particularly prominent. At the same time, increased volatility in energy prices and the advancement of energy conservation and emission reduction targets have made building energy conservation and consumption reduction a key challenge for building operation and management.

[0003] In recent years, the development of artificial intelligence and Internet of Things technologies has provided new ideas for the refined management of building energy consumption. By deploying smart sensors and edge computing devices, real-time data on all dimensions of building operation can be collected. Deep learning models can extract energy consumption patterns from historical data and accurately predict future load demand. Based on this, combined with operations research optimization algorithms to generate equipment scheduling strategies, a shift from "passive response" to "proactive optimization" can be achieved.

[0004] However, existing control strategies have poor executability and low accuracy in energy consumption prediction, so there is an urgent need to build a highly executable end-to-end intelligent energy-saving system. Summary of the Invention

[0005] This invention provides a building energy consumption control method, apparatus, equipment, and storage medium to solve the problems of poor executability and low prediction accuracy of building control strategies.

[0006] In a first aspect, the present invention provides a building energy consumption control method, comprising:

[0007] The energy consumption correlation data of the current building is obtained, and the predicted energy consumption of the current building is determined by a preset neural network based on the energy consumption correlation data, wherein the energy consumption correlation data is time series data related to the energy consumption of the current building.

[0008] For different types of electrical equipment in the current building, a mixed integer linear programming model is used to generate energy-saving control strategies based on the predicted energy consumption and preset constraints of the electrical equipment. The preset constraints are different for different types of electrical equipment.

[0009] The electrical equipment is controlled based on the energy-saving control strategy and feedback data from historical energy-saving control strategies.

[0010] Secondly, the present invention provides a building energy consumption control device, comprising:

[0011] An energy consumption prediction module is used to acquire energy consumption correlation data of the current building and use a preset neural network to determine the predicted energy consumption of the current building based on the energy consumption correlation data, wherein the energy consumption correlation data is time series data related to the energy consumption of the current building.

[0012] The energy-saving control strategy generation module is used to generate energy-saving control strategies for different types of electrical equipment in the current building using a mixed integer linear programming model based on the predicted energy consumption and preset constraints of the electrical equipment. The preset constraints are different for different types of electrical equipment.

[0013] The equipment control module is used to control the electrical equipment based on the energy-saving control strategy and feedback data from historical energy-saving control strategies.

[0014] Thirdly, the present invention provides an electronic device comprising:

[0015] At least one processor;

[0016] and memory that is communicatively connected to at least one processor;

[0017] The memory stores a computer program that can be executed by at least one processor, which enables the at least one processor to perform the building energy consumption control method of the first aspect described above.

[0018] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a processor to execute the building energy consumption control method of the first aspect described above.

[0019] The building energy consumption control scheme provided by this invention acquires the energy consumption correlation data of the current building and uses a preset neural network to determine the predicted energy consumption of the current building based on the energy consumption correlation data. The energy consumption correlation data is time-series data associated with the energy consumption of the current building. For different types of electrical equipment in the current building, a mixed-integer linear programming model is used to generate energy-saving control strategies based on the predicted energy consumption of the corresponding electrical equipment and preset constraints. The preset constraints differ for different types of electrical equipment. The electrical equipment is controlled based on the energy-saving control strategies and feedback data from historical energy-saving control strategies. By adopting the above technical solution, the preset neural network accurately predicts the building's energy consumption based on time-series data. Based on the prediction results, a mixed-integer linear programming model dynamically generates energy-saving scheduling strategies for different types of electrical equipment. This not only improves the accuracy of prediction but also enhances the executability of the control strategies, constructing a highly executable end-to-end intelligent energy-saving system.

[0020] It should be understood that the description in this section is not intended to identify key or essential features of the invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a building energy consumption control method according to Embodiment 1 of the present invention;

[0023] Figure 2 This is a flowchart of a building energy consumption control method according to Embodiment 2 of the present invention;

[0024] Figure 3 This is a schematic diagram of the structure of a building energy consumption control device according to Embodiment 3 of the present invention;

[0025] Figure 4 This is a schematic diagram of the structure of an electronic device provided according to Embodiment 4 of the present invention. Detailed Implementation

[0026] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0027] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. In the description of this invention, unless otherwise stated, "a plurality of" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist; for example, A and / or B can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0028] Example 1

[0029] Figure 1 The flowchart of a building energy consumption control method provided in Embodiment 1 of the present invention is applicable to the case of building energy consumption control. The method can be executed by a building energy consumption control device, which can be implemented in hardware and / or software. The building energy consumption control device can be configured in an electronic device, which can be composed of two or more physical entities or a single physical entity.

[0030] like Figure 1 As shown, the building energy consumption control method provided in Embodiment 1 of the present invention specifically includes the following steps:

[0031] S101. Obtain the energy consumption correlation data of the current building, and use a preset neural network to determine the predicted energy consumption of the current building based on the energy consumption correlation data, wherein the energy consumption correlation data is time series data related to the energy consumption of the current building.

[0032] In this embodiment, time-series data associated with the energy consumption of the current building can be input into a preset neural network. The preset neural network can extract energy consumption features from the time-series data, thereby predicting the predicted energy consumption of the current building in the future time period.

[0033] S102. For different types of electrical equipment in the current building, use a mixed integer linear programming model to generate energy-saving control strategies based on the predicted energy consumption and preset constraints of the electrical equipment. The preset constraints are different for different types of electrical equipment.

[0034] In this embodiment, the building can be equipped with different types of electrical equipment. These different types of equipment have different power demands and usage patterns. For example, the power consumption of lighting equipment is usually time-dependent, while that of heating equipment is usually seasonal. Therefore, a mixed-integer linear programming model can be used to generate energy-saving control strategies based on the predicted energy consumption of the electrical equipment and preset constraints. The preset constraints are used to characterize the power consumption patterns and conditions of the electrical equipment.

[0035] S103. Control the electrical equipment according to the energy-saving control strategy and the feedback data of the historical energy-saving control strategy.

[0036] For example, the gateway in the current building receives energy-saving control policies via the OPC UA protocol, and is directly connected to field device controllers via the BACnet MS / TP bus. The instructions of the energy-saving control policies are formatted and decomposed within the gateway, for example, converting the total load adjustment of the air conditioning system into the setpoint changes of specific devices, and automatically matching the control protocols of different manufacturers.

[0037] During the execution of the energy-saving control strategy, the gateway initiates a multi-level monitoring mechanism. For example, the basic layer collects the actual operating status of the equipment every 100ms, including detailed parameters such as compressor frequency and valve opening. The intermediate layer summarizes the measured energy consumption data every 5 minutes and performs deviation analysis with the predicted values. The application layer continuously monitors network quality and automatically switches to local caching strategy execution mode when communication is interrupted. All operational data undergoes preliminary processing at the edge side, is compressed and uploaded to the cloud platform via the MQTT protocol, and the raw data is stored in a local circular buffer on the gateway, retaining the historical records of the most recent 72 hours.

[0038] To assess the effectiveness of energy-saving control strategies, the gateway can dynamically evaluate and fine-tune them based on historical data. For example, if the actual energy saving rate of the lighting system is lower than expected, it automatically retrieves historical data from preset sensors to recalibrate control parameters. If the response delay of an appliance exceeds a threshold, the response time parameters of the group control algorithm are immediately adjusted. Simultaneously, the gateway incorporates a device health assessment model. By analyzing characteristic parameters such as motor current harmonics, it avoids devices in potentially faulty states when implementing energy-saving strategies. Historical energy-saving control strategies can be understood as the currently implemented energy-saving control strategies.

[0039] The building energy consumption control method provided in this embodiment of the invention acquires the energy consumption correlation data of the current building and uses a preset neural network to determine the predicted energy consumption of the current building based on the energy consumption correlation data. The energy consumption correlation data is time-series data associated with the energy consumption of the current building. For different types of electrical equipment in the current building, a mixed-integer linear programming model is used to generate energy-saving control strategies based on the predicted energy consumption of the electrical equipment and preset constraints. The preset constraints differ for different types of electrical equipment. The electrical equipment is controlled based on the energy-saving control strategies and feedback data from historical energy-saving control strategies. This technical solution of the present invention uses a preset neural network to accurately predict building energy consumption based on time-series data, and dynamically generates energy-saving scheduling strategies for different types of electrical equipment based on the prediction results using a mixed-integer linear programming model. This not only improves the accuracy of prediction but also enhances the executability of the control strategies, constructing a highly executable end-to-end intelligent energy-saving system.

[0040] Optionally, determining the predicted energy consumption of the current building using a preset neural network based on the energy consumption correlation data includes: inputting the energy consumption correlation data into a long short-term memory network with an attention mechanism to obtain the predicted energy consumption of the current building, wherein the preset neural network includes a long short-term memory network with an attention mechanism.

[0041] Specifically, the Attention-LSTM (Attention-Long Short-Term Memory) model combines attention mechanisms with Long Short-Term Memory (LSTM) networks to improve efficiency and accuracy in processing sequential data. The combination of attention mechanisms and LSTM enhances the model's ability to handle multivariate time-series data. The model first processes energy consumption-related data through LSTM layers to capture long-term dependencies. Then, the attention mechanism layer is applied to the LSTM output, dynamically adjusting the importance weights of different times and variables. This allows the model to focus on the variables and time points that have the greatest impact on the prediction target, thereby improving the accuracy and efficiency of the prediction and obtaining the predicted energy consumption of the current building.

[0042] Optionally, before determining the predicted energy consumption of the current building based on the energy consumption correlation data using a preset neural network, the method further includes: establishing a hyperparameter search space, wherein the hyperparameter search space includes at least one of the following: the number of LSTM layers, the number of hidden units, the number of attention heads, the dropout rate, and the learning rate of the preset neural network; determining the parameters to be optimized from the hyperparameter search space using the Latin hypercube sampling method, and determining the optimal value of the parameters to be optimized using 3-fold cross-validation and the normalized root mean square error function; and updating the preset neural network using the optimal value to obtain the optimized preset neural network.

[0043] Specifically, Bayesian optimization algorithms can be used to tune the hyperparameters of a pre-defined neural network to minimize prediction errors.

[0044] For example, if the neural network is preset to be a Long Short-Term Memory (Attention-LSTM) network model with an attention mechanism, the specific steps include:

[0045] When using the Bayesian optimization algorithm to tune the hyperparameters of the Attention-LSTM model, a hyperparameter search space is first established, including key parameters such as the number of LSTM layers, the number of hidden units, the number of attention heads, the Dropout rate, and the learning rate. The objective function is defined as the normalized root mean square error function to obtain the normalized root mean square error on the validation set. The specific calculation method is as follows:

[0046]

[0047] Where A represents the normalized root mean square error, and N represents the number of samples. Represented as the predicted value output by the model, y i Represented as the true value, y max Represented as the maximum value of the true value, y min Represented as the minimum value of the true value;

[0048] Bayesian optimization establishes a surrogate model using a Gaussian process (GP), and the specific calculation method is as follows:

[0049] p(f|D)~GP(μ(X),k(X,X ′ ))

[0050] Here, p(f|D) represents the posterior probability distribution of the objective function f given a dataset D. The input X in p(fD) comes from the hyperparameter search space. The objective function f is the function that maps the hyperparameter combination X to a scalar value of model performance, usually the loss on the validation set. Here, the objective function refers to the normalized root mean square error function mentioned above. D represents the dataset, and μ(X) represents the mean function, indicating the mean of a Gaussian process. k(X,X) ′ ) represents the kernel function, and represents the covariance function of the Gaussian process;

[0051] EI(x)=E[max(f min -f(x),0)]

[0052] Where EI(x) represents the expected improvement value, E(·) represents the expectation, and f min Let f(x) represent the minimum value of the objective function known at the current point x, where x represents the value of the objective function at the input point x, and x represents the input variable, i.e. the parameter to be optimized.

[0053] The optimization process is executed in two phases: the first 20 iterations use Latin hypercube sampling to search the hyperparameter space and determine the parameters to be optimized; the subsequent 80 iterations dynamically adjust the search direction based on the surrogate model. Details of the dynamic adjustment strategy include the trade-off between mean and variance, local optimization and global search, and constraint handling. In the trade-off between mean and variance, an exploration coefficient is set to adjust the exploration tendency. In local optimization and global search, multi-starting-point gradient optimization is performed on the EI function to avoid getting trapped in local optima. Every few iterations, random sampling is performed with a certain probability to ensure global coverage. In constraint handling, if there is a dependency between hyperparameters (e.g., Dropout rate ≤ 0.5), constraints are added during EI optimization.

[0054] Each trial includes 3-fold cross-validation, and the parameter combination that minimizes the normalized root mean square error of the validation set is ultimately selected. To avoid local optima, a random perturbation term is introduced in the later stages of optimization, accepting a suboptimal solution with a probability of 0.2. The optimized hyperparameters are updated through backpropagation to update the model weights. The specific calculation method is as follows:

[0055]

[0056] Here, θ* represents the optimized model weights, which is a global parameter set encompassing all trainable parameters of the attention mechanism layer, LSTM layer, and output layer. θ represents the vector of model weights, L(θ) represents the loss function, and τ represents the regularization coefficient. It is expressed as the square of the L2 norm of θ, which is the square of the Euclidean distance of the weight vector. The regularization coefficient is also included as an optimization variable in the search space.

[0057] Optionally, before determining the predicted energy consumption of the current building based on the energy consumption correlation data using a preset neural network, the method further includes: preprocessing the energy consumption correlation data to obtain processed energy consumption correlation data, wherein the energy consumption correlation data includes historical energy consumption data, environmental data, and equipment operation status data of electrical equipment, and the preprocessing includes at least one of missing value imputation, outlier removal, and feature normalization.

[0058] Specifically, historical energy consumption data can be collected primarily through metering devices such as smart meters, water meters, and gas meters. These devices automatically record electricity, water, and gas consumption data at fixed time intervals and transmit them to a data acquisition gateway via the Modbus communication protocol. For traditional meters without automatic data collection capabilities, manual meter reading can be used to supplement the data, ensuring that the timestamps are aligned with the automatically collected data. Environmental data can be collected through deployed IoT devices such as temperature and humidity sensors, concentration sensors, and illuminance sensors. These devices monitor indoor and outdoor environmental conditions at 5 to 10-minute intervals, and sensor data is uploaded to a cloud server via wireless technologies such as Zigbee or NB-IoT. Equipment operation status data comes from building equipment controllers. Start / stop signals, operating modes, and fault alarm information of air conditioning units, lighting systems, and elevator equipment are obtained through protocols such as OPC UA and BACnet, with a sampling frequency typically ranging from 1 to 5 minutes. All data is tagged with timestamps and measurement point IDs and pushed to a time-series database in real time using MQTT or HTTP protocols, while a breakpoint resume mechanism is set up to ensure data integrity in the event of network interruptions.

[0059] Then, to address the missing value issue in energy consumption-related data, we can analyze the missing data patterns. For short-term, randomly missing data, linear interpolation or forward imputation can be used to complete the data. If the data is periodic, sliding window mean or seasonal decomposition interpolation can be used for repair. For long-term breakpoints, we need to combine equipment logs to determine whether they are caused by downtime, and use a marking method instead of blind imputation. Outliers in historical energy consumption data can be removed based on the 3σ principle. For equipment operating status data, erroneous records can be identified through logical rules, and for environmental data, anomalies can be detected using isolated forests. The missing values ​​after removal are filled according to the missing value processing procedure. Finally, feature normalization is performed according to the data characteristics: energy consumption values ​​are scaled to the [0,1] interval using Min-Max. Sensor data of different dimensions are standardized using Z-Score. Categorical variables in equipment status data are converted to One-Hot encoding. The normalization parameters need to be saved for online real-time data processing. All preprocessing operations retain copies of the original data, and processing logs are recorded in the metadata to ensure data traceability.

[0060] Example 2

[0061] Figure 2 This is a flowchart of a building energy consumption control method provided in Embodiment 2 of the present invention. The technical solution of the present invention is further optimized based on the above optional technical solutions, and a specific method for building energy consumption control is given.

[0062] Optionally, the step of inputting the energy consumption-related data into a long short-term memory network with an attention mechanism to obtain the predicted energy consumption of the current building includes: processing the energy consumption-related data using a one-dimensional convolutional layer in the long short-term memory network with an attention mechanism to obtain local trend features, and using a sliding window to determine the statistical data of the energy consumption-related data, wherein the statistical data includes mean and variance time-series features; inputting the local trend features and the statistical data into a long short-term memory network layer in the long short-term memory network with an attention mechanism to obtain energy consumption cycle features with a preset time granularity; inputting the energy consumption cycle features into an attention mechanism layer in the long short-term memory network with an attention mechanism to obtain an output result, and using at least two parallel fully connected layers to process the output result respectively to obtain the load value and load change trend within a future time period; and using a weighted fusion layer to output the predicted energy consumption of the current building based on the load value and the load change trend.

[0063] Optionally, the predicted energy consumption is a predicted energy consumption curve; the predicted energy consumption curve includes the corresponding points of predicted energy consumption and time, the probability distribution of the corresponding points, and the peak energy time and the corresponding energy consumption power.

[0064] Optionally, the step of generating energy-saving control strategies for different types of electrical equipment in the current building using a mixed-integer linear programming model based on the predicted energy consumption and preset constraints of the electrical equipment includes: for air conditioners in the current building, using a mixed-integer linear programming model, with the goal of minimizing energy consumption, generating a first energy-saving control strategy based on the predicted energy consumption of the air conditioner and the temperature constraints of the area where it is located, wherein the types of electrical equipment in the current building include air conditioners, lighting equipment, and elevators; for lighting equipment, using a mixed-integer linear programming model, with the goal of minimizing energy consumption, generating a second energy-saving control strategy based on the predicted energy consumption of the lighting equipment, the light constraints of the area where it is located, and the personnel density; for elevators, using a mixed-integer linear programming model and a group control optimization algorithm, with the goal of minimizing energy consumption, generating a third energy-saving control strategy based on the predicted energy consumption of the elevator, the elevator operation quantity constraints of the area where it is located, and the personnel density.

[0065] like Figure 2 As shown in Embodiment 2 of the present invention, a building energy consumption control method specifically includes the following steps:

[0066] S201. Obtain the energy consumption correlation data of the current building, and preprocess the energy consumption correlation data to obtain processed energy consumption correlation data.

[0067] The energy consumption-related data includes historical energy consumption data, environmental data, and equipment operation status data of electrical equipment. The preprocessing includes at least one of missing value filling, outlier removal, and feature normalization.

[0068] S202. The energy consumption-related data is processed by a one-dimensional convolutional layer in a long short-term memory network with an attention mechanism to obtain local trend features, and the statistical data of the energy consumption-related data is determined by a sliding window, wherein the statistical data includes mean and variance time-series features.

[0069] S203. Input the local trend features and the statistical data into the long short-term memory network layer of the attention mechanism to obtain the energy consumption cycle features with a preset time granularity.

[0070] Specifically, a dual-channel input structure can be designed, allowing energy consumption-related data to be input into a one-dimensional convolutional layer. After the one-dimensional convolutional layer extracts local trend features, these features are then input into the model along with the mean and variance time-series features generated through sliding window statistics. The LSTM (Long Short-Term Memory) layers adopt a stacked structure. The bottom LSTM units can capture hourly fluctuation patterns and have 128 hidden nodes, while the top LSTM units can learn daily periodic characteristics. Each layer is followed by a Dropout layer to prevent overfitting, thus obtaining hourly energy consumption periodic features.

[0071] S204. Input the energy consumption cycle characteristics into the attention mechanism layer of the long short-term memory network of the attention mechanism to obtain the output result, and process the output result using at least two parallel fully connected layers to obtain the load value and load change trend in the future time period.

[0072] S205. Using the weighted fusion layer, output the predicted energy consumption curve of the current building based on the load value and the load change trend. The predicted energy consumption curve includes the corresponding points of predicted energy consumption and time, the probability distribution of the corresponding points, and the peak energy time and corresponding energy consumption power.

[0073] Specifically, the attention mechanism is deployed on the LSTM output sequence, and the attention weights for each time step are calculated using a trainable query-key matrix, focusing on key nodes during peak energy consumption periods and equipment start-up and shutdown events. A time decay factor is introduced into the weight allocation module to automatically give higher attention to recent data; the calculation method is as follows:

[0074]

[0075] Where, α t Represented as a time-weighted coefficient, W q Represented as a query vector, h tRepresented as a historical data vector, Let λ represent the scaling factor, t represent the learnable time decay coefficient, and T represent the timestamp, indicating the current time or a point in time in historical data. The attention mechanism layer output can be connected to two fully connected layers in parallel to predict the load value and load change trend for the next 24 hours, respectively. Finally, a dynamic weighted fusion layer generates the final prediction result. The training process employs an adaptive moment estimation optimizer with an initial learning rate of 0.0003, and applies a warm-start learning rate scheduling strategy, stopping early on the validation set to prevent overfitting.

[0076] in:

[0077] A Seq2Seq decoder can be used to gradually generate the prediction sequence for the next 24 hours. The calculation method is as follows:

[0078]

[0079] in, Let W represent the predicted value for the future time step generated by the decoder. o This is represented as the output weight matrix of the decoder. Represented as the hidden state of the decoder, c t Represented as a context vector, it represents the encoder's encoding of global information about the input sequence; b o This is represented as the output bias of the decoder. This is represented as a vector concatenation operation. A real-time residual feedback mechanism is introduced: when the latest actual energy consumption deviates from the predicted value by more than 15%, the LSTM hidden state is corrected.

[0080]

[0081] Among them, h t Let W represent the hidden state of the LSTM, tanh represent the activation function, and W represent the hidden state of the LSTM. r Represented as the residual feedback weight matrix, y t Represented as actual observation data, This represents the model's prediction results.

[0082] For example, the output predicted energy consumption curve for the current building may include time-of-day energy consumption predictions for the next 24 hours, specifically including:

[0083] 1) Baseline predicted value: Point estimation result That is, the predicted energy consumption value at time t+Δ;

[0084] 2) Uncertainty range: The probability distribution p(y) is generated using Monte Carlo Dropout. t+Δ|x);

[0085] 2) Key time point marking: Automatically marked peak predicted energy consumption times and their power values.

[0086] S206. For the air conditioners in the current building, a mixed integer linear programming model is used to generate a first energy-saving control strategy with the goal of minimizing energy consumption, based on the predicted energy consumption of the air conditioners and the temperature constraints corresponding to the area where they are located. The types of electrical equipment in the current building include air conditioners, lighting equipment and elevators.

[0087] S207. For the lighting equipment, using a mixed integer linear programming model, with the goal of minimizing energy consumption, a second energy-saving control strategy is generated based on the predicted energy consumption of the lighting equipment, the illumination constraints of the area, and the population density.

[0088] S208. For the elevator, using a mixed-integer linear programming model and a group control optimization algorithm, with the goal of minimizing energy consumption, a third energy-saving control strategy is generated based on the predicted energy consumption of the elevator, the constraints on the number of elevators in operation in the area, and the personnel density.

[0089] For example, if the predicted energy consumption curve is the energy consumption curve for the next 24 hours, then based on the predicted energy consumption curve for the next 24 hours, the electricity consumption period can first be divided into three price ranges: peak, flat, and valley, and the predicted peak load period can be identified.

[0090] For the air conditioning system, the thermal inertia parameters of the building envelope are pre-calculated. The energy-saving control strategies output by the mixed-integer linear programming model include: pre-cooling or preheating during off-peak electricity periods to utilize the building's heat storage characteristics to reduce the cooling load during peak hours. For temperature-sensitive areas, strict environmental parameter constraints are set to ensure that temperature and humidity remain within the set range throughout the optimization process. For general areas such as office areas, the set temperature is allowed to be dynamically adjusted within the comfort standard range, fluctuating by 2°C to reduce energy consumption.

[0091] For lighting equipment, the energy-saving control strategies output by the mixed-integer linear programming model include: dynamically adjusting the lighting brightness of different areas based on illumination prediction data, reducing artificial lighting power in areas near windows during the day, and automatically turning off unoccupied areas by combining data from personnel density sensors.

[0092] For elevators, in addition to the mixed-integer linear programming model, a group control optimization algorithm can also be used. The output energy-saving control strategies include: reducing the number of elevators in operation during off-peak hours, and predicting the passenger flow distribution of each time period based on historical usage data, and stopping some elevators at high-frequency floors in advance to reduce idle running time.

[0093] The mixed-integer linear programming model also considers the physical constraints of equipment start-up and shutdown. Each optimization generates a 15-minute granularity set of equipment control instructions, while retaining a 10% power regulation margin to address deviations between actual load and forecasts. When real-time monitoring detects that energy consumption deviates from the forecast value by more than a preset threshold, an online re-optimization mechanism is triggered to dynamically adjust the control strategy for subsequent time periods. The energy-saving control strategy can be distributed to each equipment controller for execution through the standard protocol of the building automation system, while simultaneously generating a visualized scheduling Gantt chart and an expected cost saving report for management review.

[0094] S209. Control the electrical equipment according to the energy-saving control strategy and the feedback data of the historical energy-saving control strategy.

[0095] Specifically, the edge gateway can control the air conditioning, lighting, and elevators according to the first energy-saving control strategy, the second energy-saving control strategy, the third energy-saving control strategy, and the corresponding historical energy-saving control strategies.

[0096] The building energy consumption control method provided in this invention first collects multi-source energy consumption correlation data, then uses a long short-term memory network with an attention mechanism to predict energy consumption, and then uses a mixed-integer linear programming model to dynamically generate energy-saving control strategies for air conditioning, lighting, and elevators. Finally, the energy-saving control strategies are used to implement strategy distribution and execution monitoring. This method improves prediction accuracy and strategy robustness through multi-source data acquisition, dynamic strategy generation, real-time strategy correction, and closed-loop feedback control.

[0097] Example 3

[0098] Figure 3 This is a schematic diagram of a building energy consumption control device provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes: an energy consumption prediction module 301, an energy-saving control strategy generation module 302, and an equipment control module 303, wherein:

[0099] An energy consumption prediction module is used to acquire energy consumption correlation data of the current building and use a preset neural network to determine the predicted energy consumption of the current building based on the energy consumption correlation data, wherein the energy consumption correlation data is time series data related to the energy consumption of the current building.

[0100] The energy-saving control strategy generation module is used to generate energy-saving control strategies for different types of electrical equipment in the current building using a mixed integer linear programming model based on the predicted energy consumption and preset constraints of the electrical equipment. The preset constraints are different for different types of electrical equipment.

[0101] The equipment control module is used to control the electrical equipment based on the energy-saving control strategy and feedback data from historical energy-saving control strategies.

[0102] The building energy consumption control device provided in this embodiment of the invention uses a preset neural network to accurately predict building energy consumption based on time-series data, and uses a mixed integer linear programming model based on the prediction results to dynamically generate energy-saving scheduling strategies for different types of electrical equipment. This not only improves the accuracy of prediction, but also enhances the executability of the control strategy, thus constructing a highly executable end-to-end intelligent energy-saving system.

[0103] Optional, the energy consumption prediction module includes:

[0104] An energy consumption prediction unit is used to input the energy consumption-related data into a long short-term memory network with an attention mechanism to obtain the predicted energy consumption of the current building. The preset neural network includes a long short-term memory network with an attention mechanism.

[0105] Furthermore, the step of inputting the energy consumption-related data into a long short-term memory network with an attention mechanism to obtain the predicted energy consumption of the current building includes: processing the energy consumption-related data using a one-dimensional convolutional layer in the long short-term memory network with an attention mechanism to obtain local trend features, and using a sliding window to determine the statistical data of the energy consumption-related data, wherein the statistical data includes mean and variance time-series features; inputting the local trend features and the statistical data into a long short-term memory network layer in the long short-term memory network with an attention mechanism to obtain energy consumption cycle features with a preset time granularity; inputting the energy consumption cycle features into an attention mechanism layer in the long short-term memory network with an attention mechanism to obtain an output result, and using at least two parallel fully connected layers to process the output result respectively to obtain the load value and load change trend within a future time period; and using a weighted fusion layer to output the predicted energy consumption of the current building based on the load value and the load change trend.

[0106] Furthermore, the predicted energy consumption is a predicted energy consumption curve; the predicted energy consumption curve includes the corresponding points of predicted energy consumption and time, the probability distribution of the corresponding points, and the peak energy time and the corresponding energy consumption power.

[0107] Optionally, the device may also include:

[0108] The space establishment module is used to establish a hyperparameter search space before determining the predicted energy consumption of the current building based on the energy consumption correlation data using a preset neural network. The hyperparameter search space includes at least one of the following: the number of LSTM layers, the number of hidden units, the number of attention heads, the Dropout rate, and the learning rate of the preset neural network.

[0109] The optimal value determination module is used to determine the parameter to be optimized from the hyperparameter search space according to the Latin hypercube sampling method, and to determine the optimal value of the parameter to be optimized through 3-fold cross-validation and normalized root mean square error function.

[0110] An update module is used to update the preset neural network using the optimal value to obtain an optimized preset neural network.

[0111] Optionally, the energy-saving control strategy generation module includes:

[0112] The first energy-saving control strategy generation unit is used to generate a first energy-saving control strategy for the air conditioners in the current building using a mixed integer linear programming model with the goal of minimizing energy consumption, based on the predicted energy consumption of the air conditioners and the temperature constraints corresponding to the area where they are located. The types of electrical equipment in the current building include air conditioners, lighting equipment and elevators.

[0113] The second energy-saving control strategy generation unit is used to generate a second energy-saving control strategy for the lighting equipment by using a mixed integer linear programming model with the goal of minimizing energy consumption, based on the predicted energy consumption of the lighting equipment, the light constraints corresponding to the area where it is located, and the population density.

[0114] The third energy-saving control strategy generation unit is used to generate a third energy-saving control strategy for the elevator, using a mixed integer linear programming model and a group control optimization algorithm, with the goal of minimizing energy consumption, based on the predicted energy consumption of the elevator, the constraints on the number of elevators in operation in the area, and the personnel density.

[0115] Optionally, the device may also include:

[0116] The preprocessing module is used to preprocess the energy consumption association data before determining the predicted energy consumption of the current building based on the energy consumption association data using a preset neural network, so as to obtain processed energy consumption association data. The energy consumption association data includes historical energy consumption data, environmental data, and equipment operation status data of electrical equipment. The preprocessing includes at least one of missing value imputation, outlier removal, and feature normalization.

[0117] The building energy consumption control device provided in the embodiments of the present invention can execute the building energy consumption control method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0118] Example 4

[0119] Figure 4A schematic diagram of an electronic device 40 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0120] like Figure 4 As shown, the electronic device 40 includes at least one processor 41 and a memory, such as a read-only memory (ROM) 42 or a random access memory (RAM) 43, communicatively connected to the at least one processor 41. The memory stores computer programs executable by the at least one processor. The processor 41 can perform various appropriate actions and processes based on the computer program stored in the ROM 42 or loaded into the RAM 43 from storage unit 48. The RAM 43 may also store various programs and data required for the operation of the electronic device 40. The processor 41, ROM 42, and RAM 43 are interconnected via a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0121] Multiple components in electronic device 40 are connected to I / O interface 45, including: input unit 46, such as keyboard, mouse, etc.; output unit 47, such as various types of monitors, speakers, etc.; storage unit 48, such as disk, optical disk, etc.; and communication unit 49, such as network card, modem, wireless transceiver, etc. Communication unit 49 allows electronic device 40 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0122] Processor 41 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 41 performs the various methods and processes described above, such as building energy consumption control methods.

[0123] In some embodiments, the building energy consumption control method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 48. In some embodiments, part or all of the computer program may be loaded into and / or installed on electronic device 40 via ROM 42 and / or communication unit 49. When the computer program is loaded into RAM 43 and executed by processor 41, one or more steps of the building energy consumption control method described above may be performed. Alternatively, in other embodiments, processor 41 may be configured to perform the building energy consumption control method by any other suitable means (e.g., by means of firmware).

[0124] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoC) systems, complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

[0125] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0126] The computer equipment provided above can be used to execute the building energy consumption control method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0127] Example 5

[0128] In the context of this invention, the computer-readable storage medium may be a tangible medium, and the computer-executable instructions, when executed by a computer processor, are used to perform a building energy consumption control method, the method comprising:

[0129] The energy consumption correlation data of the current building is obtained, and the predicted energy consumption of the current building is determined by a preset neural network based on the energy consumption correlation data, wherein the energy consumption correlation data is time series data related to the energy consumption of the current building.

[0130] For different types of electrical equipment in the current building, a mixed integer linear programming model is used to generate energy-saving control strategies based on the predicted energy consumption and preset constraints of the electrical equipment. The preset constraints are different for different types of electrical equipment.

[0131] The electrical equipment is controlled based on the energy-saving control strategy and feedback data from historical energy-saving control strategies.

[0132] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by, or in conjunction with, an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0133] The computer equipment provided above can be used to execute the building energy consumption control method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0134] It is worth noting that in the above-mentioned embodiments of building energy consumption control devices, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0135] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A building energy consumption control method, characterized in that, include: The energy consumption correlation data of the current building is obtained, and the predicted energy consumption of the current building is determined by a preset neural network based on the energy consumption correlation data, wherein the energy consumption correlation data is time series data related to the energy consumption of the current building. For different types of electrical equipment in the current building, a mixed integer linear programming model is used to generate energy-saving control strategies based on the predicted energy consumption and preset constraints of the electrical equipment. The preset constraints are different for different types of electrical equipment. The electrical equipment is controlled based on the energy-saving control strategy and feedback data from historical energy-saving control strategies.

2. The method according to claim 1, characterized in that, The step of determining the predicted energy consumption of the current building using a preset neural network based on the energy consumption correlation data includes: The energy consumption-related data is input into a long short-term memory network with an attention mechanism to obtain the predicted energy consumption of the current building. The preset neural network includes a long short-term memory network with an attention mechanism.

3. The method according to claim 2, characterized in that, The step of inputting the energy consumption-related data into a long short-term memory network with an attention mechanism to obtain the predicted energy consumption of the current building includes: The energy consumption-related data is processed by a one-dimensional convolutional layer in a long short-term memory network with an attention mechanism to obtain local trend features, and a sliding window is used to determine the statistical data of the energy consumption-related data, wherein the statistical data includes mean and variance time-series features. The local trend features and the statistical data are input into the long short-term memory network layer of the attention mechanism to obtain the energy consumption cycle features with a preset time granularity. The energy consumption cycle characteristics are input into the attention mechanism layer of the long short-term memory network of the attention mechanism to obtain the output result. The output result is then processed by at least two parallel fully connected layers to obtain the load value and load change trend in the future time period. The weighted fusion layer is used to output the predicted energy consumption of the current building based on the load value and the load change trend.

4. The method according to claim 2, characterized in that, The predicted energy consumption is a predicted energy consumption curve; the predicted energy consumption curve includes the corresponding points of predicted energy consumption and time, the probability distribution of the corresponding points, and the peak energy time and the corresponding energy consumption power.

5. The method according to any one of claims 1-4, characterized in that, Before determining the predicted energy consumption of the current building based on the energy consumption correlation data using a preset neural network, the method further includes: Establish a hyperparameter search space, wherein the hyperparameter search space includes at least one of the following: the number of LSTM layers, the number of hidden units, the number of attention heads, the Dropout rate, and the learning rate of the preset neural network; The parameters to be optimized are determined from the hyperparameter search space using the Latin hypercube sampling method, and the optimal values ​​of the parameters to be optimized are determined by 3-fold cross-validation and normalized root mean square error function. The optimal value is used to update the preset neural network to obtain the optimized preset neural network.

6. The method according to claim 1, characterized in that, The method involves generating energy-saving control strategies for different types of electrical equipment within the building, using a mixed-integer linear programming model based on the predicted energy consumption and preset constraints of the electrical equipment. These strategies include: For the air conditioners in the current building, a mixed integer linear programming model is used to generate a first energy-saving control strategy with the goal of minimizing energy consumption, based on the predicted energy consumption of the air conditioners and the temperature constraints of the area where they are located. The types of electrical equipment in the current building include air conditioners, lighting equipment and elevators. For the aforementioned lighting equipment, a mixed-integer linear programming model is used to generate a second energy-saving control strategy with the goal of minimizing energy consumption, based on the predicted energy consumption of the lighting equipment, the illumination constraints of the area, and the population density. For the elevator, a third energy-saving control strategy is generated using a mixed-integer linear programming model and a group control optimization algorithm, with the goal of minimizing energy consumption, based on the predicted energy consumption of the elevator, the constraints on the number of elevators in operation in the area, and the personnel density.

7. The method according to claim 1, characterized in that, Before determining the predicted energy consumption of the current building based on the energy consumption correlation data using a preset neural network, the method further includes: The energy consumption correlation data is preprocessed to obtain processed energy consumption correlation data, wherein the energy consumption correlation data includes historical energy consumption data, environmental data and equipment operation status data of electrical equipment, and the preprocessing includes at least one of missing value imputation, outlier removal and feature normalization.

8. A building energy consumption control device, characterized in that, include: An energy consumption prediction module is used to acquire energy consumption correlation data of the current building and use a preset neural network to determine the predicted energy consumption of the current building based on the energy consumption correlation data, wherein the energy consumption correlation data is time series data related to the energy consumption of the current building. The energy-saving control strategy generation module is used to generate energy-saving control strategies for different types of electrical equipment in the current building using a mixed integer linear programming model based on the predicted energy consumption and preset constraints of the electrical equipment. The preset constraints are different for different types of electrical equipment. The equipment control module is used to control the electrical equipment based on the energy-saving control strategy and feedback data from historical energy-saving control strategies.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the building energy consumption control method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the building energy consumption control method according to any one of claims 1-7.

Images