Green building energy-saving design method and system based on deep learning

Abstract

The invention discloses a green building energy-saving design method and system based on deep learning, and belongs to the technical field of data processing. A green building energy consumption prediction model is constructed by adopting a deep learning model, and historical sample data is used as data support; according to the method, a green building energy consumption prediction model is optimized by adopting a multi-strategy collaborative optimization algorithm to obtain an optimized green building energy consumption prediction model, and then the optimized green building energy consumption prediction model can be scheduled to analyze real-time building energy consumption so as to realize energy consumption prediction. And finally, energy-saving management can be carried out according to the predicted energy consumption data, so that the energy-saving management efficiency can be effectively improved, energy-saving management of the green building can be carried out in real time, and the design of energy-saving management by workers is facilitated.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and particularly relates to a green building energy-saving design method and system based on deep learning. Background Art

[0002] The construction industry plays a crucial role in global energy consumption and greenhouse gas emissions and is a key area for achieving energy conservation and emission reduction goals. With the accelerating urbanization rate, the electricity consumption and carbon emissions of residential buildings have increased sharply. Accurately predicting and analyzing the electricity consumption of residential buildings and then formulating effective energy conservation and emission reduction strategies are of great significance for energy conservation and emission reduction. With the rapid development of China's economy, the problem of building energy consumption has become increasingly serious. Green building energy-saving design has become an important means to reduce building energy consumption and improve the building environment quality. However, the traditional building energy-saving design methods have the following deficiencies: long design cycle and low efficiency; strong dependence on designers' experience, resulting in uneven design quality; and difficulty in realizing real-time optimization of building energy consumption and the environment. Summary of the Invention

[0003] The present invention provides a green building energy-saving design method and system based on deep learning to solve the problems of low efficiency of energy-saving design and difficulty in realizing real-time optimization in the prior art.

[0004] In a first aspect, the present invention provides a green building energy-saving design method based on deep learning, including: Collecting historical building energy consumption characteristics corresponding to a target green building and historical energy consumption data corresponding to the historical building energy consumption characteristics, and obtaining historical sample data after preprocessing the historical building energy consumption characteristics and the historical energy consumption data corresponding to the historical building energy consumption characteristics; Constructing a green building energy consumption prediction model by using a deep learning model, and optimizing the green building energy consumption prediction model by using a multi-strategy collaborative optimization algorithm with the historical sample data as data support to obtain an optimized green building energy consumption prediction model; Collecting real-time building energy consumption characteristics corresponding to the target green building, scheduling the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics, determining predicted energy consumption data, and performing energy-saving management of the green building according to the predicted energy consumption data.

[0005] Further, collecting historical building energy consumption characteristics corresponding to a target green building and historical energy consumption data corresponding to the historical building energy consumption characteristics, and obtaining historical sample data after preprocessing the historical building energy consumption characteristics and the historical energy consumption data corresponding to the historical building energy consumption characteristics, including: Collect the average daily temperature, daily maximum temperature, daily minimum temperature, date type, season type, and weather of the target green building on any day to obtain the historical building energy consumption characteristics corresponding to the green building; For the historical building energy consumption characteristics corresponding to any green building, collect the historical energy consumption data corresponding to the historical building energy consumption characteristics; After normalizing the historical building energy consumption characteristics, obtain the normalized historical building energy consumption characteristics, construct the normalized historical building energy consumption characteristics into input samples, and construct the historical energy consumption data corresponding to the historical building energy consumption characteristics into expected output labels to obtain historical sample data.

[0006] Furthermore, construct a green building energy consumption prediction model using a deep learning model, including: constructing a green building energy consumption prediction model using a CNN-LSTM model.

[0007] Furthermore, using the historical sample data as data support, optimize the green building energy consumption prediction model using a multi-strategy collaborative optimization algorithm to obtain an optimized green building energy consumption prediction model, including: Initialize the hyperparameters of the green building energy consumption prediction model using a chaotic mapping initialization method or a random initialization method, encode the hyperparameters to obtain an encoded vector, and obtain multiple different encoded vectors; Using the input samples in the historical sample data as the actual input and the expected output labels in the historical sample data as the expected output, obtain the fitness corresponding to each encoded vector; According to the order of the fitness corresponding to the encoded vectors from largest to smallest, divide all the encoded vectors into an optimal region search population, a collaborative balance search population, and a fast global search population in turn; For any first target encoded vector in the optimal region search population, search the first target encoded vector using a boundary probability transfer strategy to obtain the searched first target encoded vector; For any second target encoded vector in the collaborative balance search population, search the second target encoded vector using a spiral balance search strategy to obtain the searched second target encoded vector; For any third target encoded vector in the fast global search population, search the third target encoded vector using an optimal attraction and worst warning search strategy to obtain the searched third target encoded vector; Determine whether the optimization end condition is satisfied. If so, determine the optimized green building energy consumption prediction model according to the searched first target encoded vector, the searched second target encoded vector, and the searched third target encoded vector. Otherwise, return to the step of determining the fitness corresponding to each encoded vector.

[0008] Further, for any first target coding vector in the best region search population, the boundary probability transfer strategy is adopted to search the first target coding vector, and the first target coding vector after the search is obtained, including: For any first target coding vector in the best region search population, based on the boundaries of each dimension hyperparameter, the boundary transfer value corresponding to the first target coding vector is obtained as:

[0009] where, represents the t th optimization process, the j th dimension hyperparameter corresponding to the th first target coding vector, i represents the j th dimension hyperparameter corresponding to the th first target coding vector, i = 1, 2, …, L, where L represents the total number of first target coding vectors, j = 1, 2, …, K, where K represents the total dimension of hyperparameters, represents the upper limit of the th dimension hyperparameter, represents the lower limit of the th dimension hyperparameter, represents the j th dimension hyperparameter of the optimal coding vector; For any first target coding vector in the best region search population, the neighborhood search value of the first target coding vector is obtained as:

[0010] where, represents the neighborhood search value of the first target coding vector, represents the th dimension hyperparameter corresponding to a randomly selected other first target coding vector, represents the circumference ratio, represents the preset maximum number of optimizations, cos represents the cosine function, represents the third random number between (0, 0.55); Generate the first decision factor as: ; where, represents the first decision factor, represents the control hyperparameter corresponding to the first decision factor and is set to 0.499; Generate a fourth random number between (0, 1) and determine whether the fourth random number is less than the first decision factor . If so, use the neighborhood search value as the first target coding vector after the search; otherwise, judge the fitness of the boundary transfer value; Judge whether the fitness value of the boundary transfer value is greater than the fitness corresponding to the corresponding first target coding vector. If so, use the boundary transfer value as the first target coding vector after the search; otherwise, directly use the original first target coding vector as the first target coding vector after the search.

[0011] Furthermore, for any second target coding vector in the collaborative balance search population, use the spiral balance search strategy to search for the second target coding vector to obtain the second target coding vector after the search, including: For any second target coding vector in the collaborative balance search population, obtain the fitness corresponding to the second target coding vector; Arrange the second target coding vectors in descending order according to the fitness corresponding to the second target coding vectors to obtain the second target coding vectors after the arrangement; Based on the second target coding vectors after the arrangement, for the m-th second target coding vector, search for the second target coding vector to obtain the second target coding vector after the search as:

[0012] wherein, represents the t -th optimization process, the m -th hyperparameter of the second target coding vector after the arrangement, j dimension, j = 1, 2, …, K, K [[ID=...]] (the remaining content is translated in the same way as above) represents the total dimension of the hyperparameters, m = M / 2 + 1, M / 2 + 2, …, M, M represents the total number of second target coding vectors, and M is an even number; represents the m -th hyperparameter of the second target coding vector after the search, j dimension, represents the natural constant function, represents the spiral search control factor, and ; represents the variation coefficient and is set to 5; represents the pi, represents the cosine function, represents the exponential function with the natural constant e as the base, represents the preset maximum number of optimizations, A random spiral shape control factor uniformly distributed between [-1, 1], and abs represents the absolute value function. represents the j dimensional hyperparameter of the worst encoded vector among all encoded vectors; Based on the second target encoded vector after permutation, for the g-th second target encoded vector, search the second target encoded vector to obtain the second target encoded vector after search as:

[0013] where represents the j dimensional hyperparameter of the optimal encoded vector among all encoded vectors, g = 1, 2, …, M / 2, represents the t dimensional hyperparameter of the g th permuted second target encoded vector in the j th optimization process, represents the g dimensional hyperparameter of the j th second target encoded vector after search, represents a direction search factor with a random dimension of -1 or 1.

[0014] Furthermore, for any third target encoded vector in the fast global search population, use the optimal attraction and worst warning search strategy to search the third target encoded vector to obtain the third target encoded vector after search, including: Randomly generate a fifth random number between (0, 1) ; Judge whether the fifth random number is less than the preset second decision factor , if so, search the third target encoded vector according to the optimal parameter encoding to obtain the third target encoded vector after search, otherwise search the third target encoded vector according to the worst parameter encoding to obtain the third target encoded vector after search; where the second decision factor is set to t / , represents the maximum number of optimizations; Search the third target encoded vector according to the optimal parameter encoding to obtain the third target encoded vector after search as:

[0015] where represents the t dimensional hyperparameter of the k th third target encoded vector in thej Dimensional hyperparameter, k = 1, 2, …, P, where P represents the total number of third target coding vectors, j = 1, 2, …, K, K represents the total dimension of hyperparameters, represents the k th dimensional hyperparameter of the third target coding vector after the j th search, represents the j th dimensional hyperparameter of the optimal coding vector among all coding vectors, represents the sixth random number between (0, 1), represents the seventh random number between (0, 1), represents the sine function, represents the first position weighting factor, represents the second position weighting factor; Search for the third target coding vector according to the worst parameter coding, and the third target coding vector after the search is:

[0016] where, represents the eighth random number between (0, 1), represents the j th dimensional hyperparameter of the worst coding vector among all coding vectors, represents the t th fitness corresponding to the k th third target coding vector in the th optimization process, represents the fitness of the worst coding vector,

[0017] Furthermore, according to the first target coding vector after the search, the second target coding vector after the search, and the third target coding vector after the search, determine the optimized green building energy consumption prediction model, including: Re-obtain the optimal coding vector according to the first target coding vector after the search, the second target coding vector after the search, and the third target coding vector after the search; Use the hyperparameters included in the re-obtained optimal coding vector as the final hyperparameters of the green building energy consumption prediction model to obtain the optimized green building energy consumption prediction model.

[0018] Furthermore, collect the real-time building energy consumption characteristics of the target green building, and schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics, determine the predicted energy consumption data, and perform energy-saving management of the green building according to the predicted energy consumption data, including: Collect the real-time building energy consumption characteristics corresponding to the target green building, and schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics to determine the predicted energy consumption data; Based on multiple preset energy consumption data intervals, determine the energy consumption data interval to which the predicted energy consumption data belongs to obtain the target energy consumption data interval; Schedule the target energy-saving management plan corresponding to the target energy consumption data interval, and execute the target energy-saving management plan to achieve the energy-saving management of the green building; wherein, each energy consumption data interval is preset with an energy-saving management plan.

[0019] In a second aspect, the present invention provides a green building energy-saving design system based on deep learning, including: a historical data acquisition module, a deep learning module, and a building energy-saving management module; The historical data acquisition module is used to collect the historical building energy consumption characteristics corresponding to the target green building and the historical energy consumption data corresponding to the historical building energy consumption characteristics, and after preprocessing the historical building energy consumption characteristics and the historical energy consumption data corresponding to the historical building energy consumption characteristics, obtain historical sample data; The deep learning module is used to construct a green building energy consumption prediction model by using a deep learning model, and use the historical sample data as data support to optimize the green building energy consumption prediction model by using a multi-strategy collaborative optimization algorithm to obtain an optimized green building energy consumption prediction model; The building energy-saving management module is used to collect the real-time building energy consumption characteristics corresponding to the target green building, and schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics to determine the predicted energy consumption data, and perform energy-saving management of the green building according to the predicted energy consumption data.

[0020] A green building energy-saving design method and system provided by the present invention, by using a deep learning model to construct a green building energy consumption prediction model, using the historical sample data as data support, and optimizing the green building energy consumption prediction model by using a multi-strategy collaborative optimization algorithm to obtain an optimized green building energy consumption prediction model, and then can schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics to achieve energy consumption prediction, and finally can perform energy-saving management according to the predicted energy consumption data, which can not only effectively improve the energy-saving management efficiency and perform real-time energy-saving management of green buildings, but also be more conducive to the design of energy-saving management by staff. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present invention, and are used together with the specification to explain the principles of the present invention.

[0022] Figure 1 A flowchart of a green building energy-saving design method based on deep learning provided by an embodiment of the present invention.

[0023] Figure 2 A schematic structural diagram of a green building energy-saving design system based on deep learning provided by an embodiment of the present invention.

[0024] Among them, 21 - historical data acquisition module, 22 - deep learning module, 23 - building energy-saving management module.

[0025] Through the above-mentioned drawings, specific embodiments of the present invention have been shown, and there will be more detailed descriptions hereinafter. These drawings and text descriptions are not intended to limit the scope of the inventive concept in any way, but to illustrate the concept of the present invention to those skilled in the art by referring to specific embodiments. Detailed Description of Specific Embodiments

[0026] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0027] The embodiments of the present invention will be described in detail below with reference to the drawings.

[0028] As Figure 1 shown, an embodiment of the present invention provides a green building energy-saving design method based on deep learning, including: S11. Collect the historical building energy consumption characteristics corresponding to the target green building and the historical energy consumption data corresponding to the historical building energy consumption characteristics, and after preprocessing the historical building energy consumption characteristics and the historical energy consumption data corresponding to the historical building energy consumption characteristics, obtain historical sample data; The historical building energy consumption characteristics corresponding to the target green building refer to some characteristics related to energy consumption. The historical building energy consumption characteristics and the historical energy consumption data corresponding to the historical building energy consumption characteristics can be collected simultaneously. After learning the collected data, prediction can be realized.

[0029] S12. Use a deep learning model to construct a green building energy consumption prediction model, and use the historical sample data as data support, and optimize the green building energy consumption prediction model by using a multi-strategy collaborative optimization algorithm to obtain an optimized green building energy consumption prediction model; A deep learning model with data prediction capabilities can be used to construct a green building energy consumption prediction model. After constructing the green building energy consumption prediction model, it is necessary to optimize the hyperparameters of the green building energy consumption prediction model in order to achieve data recognition. In the prior art, the gradient descent method is commonly used for hyperparameter optimization, resulting in poor optimization ability and thus poor energy consumption prediction effect. Therefore, the embodiments of the present invention use a multi-strategy collaborative optimization algorithm to optimize the green building energy consumption prediction model to improve the optimization speed and global optimization ability of the algorithm.

[0030] S13. Collect the real-time building energy consumption characteristics corresponding to the target green building, and schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics, determine the predicted energy consumption data, and perform energy conservation management of the green building according to the predicted energy consumption data.

[0031] Performing energy conservation management of the green building according to the predicted energy consumption data may be: performing energy conservation management using an energy conservation management plan preset by the staff.

[0032] In the embodiments of the present invention, the historical building energy consumption characteristics corresponding to the target green building and the historical energy consumption data corresponding to the historical building energy consumption characteristics are collected, and after preprocessing the historical building energy consumption characteristics and the historical energy consumption data corresponding to the historical building energy consumption characteristics, historical sample data is obtained, including: Collect the average daily temperature, daily maximum temperature, daily minimum temperature, date type, season type, and weather of the target green building on any day to obtain the historical building energy consumption characteristics corresponding to the green building; For the historical building energy consumption characteristics corresponding to any green building, collect the historical energy consumption data corresponding to the historical building energy consumption characteristics; After normalizing the historical building energy consumption characteristics, the normalized historical building energy consumption characteristics are obtained, and the normalized historical building energy consumption characteristics are constructed as input samples, and the historical energy consumption data corresponding to the historical building energy consumption characteristics are constructed as expected output labels to obtain historical sample data.

[0033] It should be noted that the historical building energy consumption characteristics corresponding to the above green building are only the preferred embodiments provided by the embodiments of the present invention, and other characteristics may also be used as the historical building energy consumption characteristics corresponding to the green building.

[0034] In the embodiments of the present invention, a deep learning model is used to construct a green building energy consumption prediction model, including: using a CNN-LSTM (Convolutional Neural Network-Long Short-Term Memory Network) model to construct a green building energy consumption prediction model.

[0035] An embodiment of the present invention proposes a green building energy consumption prediction model based on a hybrid model of a convolutional neural network (CNN) and a long short-term memory network (LSTM), which combines the advantages of CNN in feature extraction and the ability of LSTM in time series prediction, and can effectively improve the energy consumption prediction ability.

[0036] In the embodiment of the present invention, with the historical sample data as the data support, a multi-strategy collaborative optimization algorithm is used to optimize the green building energy consumption prediction model, and the optimized green building energy consumption prediction model is obtained, including: Initialize the hyperparameters of the green building energy consumption prediction model by using a chaotic mapping initialization method or a random initialization method, encode the hyperparameters to obtain an encoded vector, and obtain a plurality of different encoded vectors; Take the input samples in the historical sample data as the actual input and the expected output labels in the historical sample data as the expected output to obtain the fitness corresponding to each encoded vector; For example, the encoded vector can be applied to the green building energy consumption prediction model, then take the input samples in the historical sample data as the actual input and the expected output labels in the historical sample data as the expected output to obtain the error function value corresponding to the encoded vector, and then add the error function value to a preset constant term (such as 0.001) and take the reciprocal, so as to obtain the fitness corresponding to the encoded vector.

[0037] Arrange all the encoded vectors in descending order according to the fitness corresponding to the encoded vectors, and divide them into the best region search population, the collaborative balance search population, and the fast global search population in turn; For any first target encoded vector in the best region search population, use a boundary probability transfer strategy to search the first target encoded vector to obtain the searched first target encoded vector; For any second target encoded vector in the collaborative balance search population, use a spiral balance search strategy to search the second target encoded vector to obtain the searched second target encoded vector; For any third target encoded vector in the fast global search population, use an optimal attraction and worst warning search strategy to search the third target encoded vector to obtain the searched third target encoded vector; Judge whether the optimization end condition is satisfied (such as the current optimization times reach the preset maximum optimization times). If so, determine the optimized green building energy consumption prediction model according to the searched first target encoded vector, the searched second target encoded vector, and the searched third target encoded vector. Otherwise, return to the step of determining the fitness corresponding to each encoded vector.

[0038] Optionally, after searching the first target coding vector, the second target coding vector, and the third target coding vector, boundary crossing processing can be performed on the first target coding vector, the second target coding vector, and the third target coding vector, so as to ensure that the algorithm is always effective.

[0039] In an embodiment of the present invention, for any first target coding vector in the best region search population, a boundary probability transfer strategy is used to search the first target coding vector, and the first target coding vector after the search is obtained, including: For any first target coding vector in the best region search population, based on the boundaries of each dimension hyperparameter, the boundary transfer value corresponding to the first target coding vector is obtained as:

[0040] Wherein, represents the t th optimization process, the j th dimension hyperparameter corresponding to the th first target coding vector, i represents the j th dimension hyperparameter corresponding to the th first target coding vector, i = 1, 2,..., L, L represents the total number of first target coding vectors, j = 1, 2,..., K, K represents the total dimension of hyperparameters, represents the upper limit of the th dimension hyperparameter, represents the lower limit of the th dimension hyperparameter, represents the first random number between (0, 1), j represents the second random number between (0, 1), represents the

[0041] th dimension hyperparameter of the optimal coding vector; For any first target coding vector in the best region search population, the neighborhood search value of the first target coding vector is obtained as: represents the neighborhood search value of the first target coding vector, represents the th dimension hyperparameter corresponding to a randomly selected other first target coding vector, represents pi, represents the preset maximum number of optimizations, cos represents the cosine function, ; wherein, represents the first decision factor, represents the control hyperparameter corresponding to the first decision factor and is set to 0.499; represents the exponential function with the natural constant e as the base; Generate a fourth random number between (0, 1) , and determine the fourth random number whether it is less than the first decision factor , if so, use the neighborhood search value as the first target coding vector after search, otherwise perform the fitness judgment of the boundary transfer value; Judge whether the fitness value of the boundary transfer value is greater than the fitness corresponding to the corresponding first target coding vector. If so, use the boundary transfer value as the first target coding vector after search, otherwise directly use the original first target coding vector as the first target coding vector after search.

[0042] In the embodiment of the present invention, the boundary probability transfer strategy is adopted to search for the first target coding vector, which can guide the first target coding vector to search in the boundary direction in the early stage of the algorithm, and adjust the change amplitude according to the optimal coding vector during the position update process, so as to prevent the problem of missing the global optimal solution caused by too fast convergence speed. In the later stage of the algorithm, it is more inclined to perform neighborhood search, which can effectively fuse the regions between two first target coding vectors, and the higher the search accuracy in the later stage of the algorithm, which can effectively improve the search accuracy of the algorithm and the ability to find the global optimal solution.

[0043] In the embodiment of the present invention, for any second target coding vector in the cooperative balance search population, the spiral balance search strategy is adopted to search for the second target coding vector to obtain the second target coding vector after search, including: For any second target coding vector in the cooperative balance search population, obtain the fitness corresponding to the second target coding vector; According to the fitness corresponding to the second target coding vector, arrange the second target coding vectors in descending order to obtain the arranged second target coding vectors; Based on the arranged second target coding vectors, for the m-th second target coding vector, search for the second target coding vector, and the second target coding vector after search is:

[0044] where represents the t -th optimization process, the m -th dimension hyperparameter of the j -th arranged second target coding vector, j = 1, 2,..., K, KIt represents the total dimension of hyperparameters, where \(m = M / 2 + 1, M / 2 + 2, \ldots, M\), \(M\) represents the total number of the second target coding vectors, and \(M\) is an even number; It represents the m th j -dimensional hyperparameter of the second target coding vector after the th search, It represents the natural constant function, ; It represents the variation coefficient and is set to 5; It represents the pi, It represents the cosine function, It represents the exponential function with the natural constant \(e\) as the base, It represents the preset maximum number of optimization times, It represents the random spiral shape control factor uniformly distributed between \([-1, 1]\), and \(abs\) represents the absolute value function, It represents the j -dimensional hyperparameter of the worst coding vector among all coding vectors; Based on the arranged second target coding vectors, for the \(g\)th second target coding vector, search the second target coding vectors, and the obtained second target coding vector after the search is:

[0045] where, It represents the j -dimensional hyperparameter of the best coding vector among all coding vectors, g \(i = 1, 2, \ldots, M / 2\), It represents the t th g -dimensional hyperparameter of the \(i\)th arranged second target coding vector during the j th optimization process, It represents the g th j -dimensional hyperparameter of the second target coding vector after the th search,

[0046] The embodiment of the present invention adopts a spiral balance search strategy to search the second target coding vectors, which can perform a larger range of searches for some second target codings with poor fitness, and perform better searches for some second target codings with better fitness. At the same time, it combines the cosine function and the exponential function, which can balance the global search and local search capabilities during the search process.

[0047] In the embodiment of the present invention, for any third target coding vector in the quickly globally searched population, the optimal attraction and worst warning search strategy is adopted to search the third target coding vector, and the third target coding vector after the search is obtained, including: Randomly generate a fifth random number between (0, 1) ; Judge the fifth random number Whether it is less than the preset second decision factor , if so, search the third target coding vector according to the optimal parameter coding to obtain the third target coding vector after the search, otherwise search the third target coding vector according to the worst parameter coding to obtain the third target coding vector after the search; where the second decision factor Is set to t / , Indicates the maximum number of optimizations; Search the third target coding vector according to the optimal parameter coding, and the third target coding vector after the search is:

[0048] Wherein, Represents the t th optimization process, the k th dimension hyperparameter of the j th third target coding vector, k = 1, 2,..., P, P represents the total number of third target coding vectors, j = 1, 2,..., K, K Represents the total dimension of hyperparameters, Represents the k th dimension hyperparameter of the j th third target coding vector after the search, Represents the j th dimension hyperparameter of the optimal coding vector among all coding vectors, Represents the sixth random number between (0, 1), Represents the seventh random number between (0, 1), Represents the sine function, Represents the first position weighting factor, Represents the second position weighting factor; Search the third target coding vector according to the worst parameter coding, and the third target coding vector after the search is:

[0049] Wherein, Represents the eighth random number between (0, 1), The j -dimensional hyperparameter representing the worst encoding vector among all encoding vectors, represents the t fitness corresponding to the k th third target encoding vector during the th optimization process, represents the fitness of the worst encoding vector,

[0050] In the embodiments of the present invention, an optimal attraction and worst alert search strategy is adopted to search for the third target encoding vector, which can search in the direction away from the worst position in the early stage of the algorithm, effectively avoid the algorithm falling into local optimum, and at the same time can be more inclined to search around the optimal area in the later stage of the algorithm, improve the search accuracy, and perform fluctuating search, which is more conducive to searching for the global optimal position.

[0051] Several search methods provided in the embodiments of the present invention cooperate with each other, can search more solution space regions, avoid search collisions, effectively avoid falling into local optimum, improve the search speed and accuracy, and finally improve the accuracy of building energy consumption prediction.

[0052] In the embodiments of the present invention, according to the first target encoding vector after search, the second target encoding vector after search, and the third target encoding vector after search, an optimized green building energy consumption prediction model is determined, including: According to the first target encoding vector after search, the second target encoding vector after search, and the third target encoding vector after search, re-obtain the optimal encoding vector; Use the hyperparameters included in the re-obtained optimal encoding vector as the final hyperparameters of the green building energy consumption prediction model to obtain the optimized green building energy consumption prediction model.

[0053] In the embodiments of the present invention, collect the real-time building energy consumption characteristics of the target green building, and schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics to determine the predicted energy consumption data, and perform energy-saving management of the green building according to the predicted energy consumption data, including: Collect the real-time building energy consumption characteristics of the target green building, and schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics to determine the predicted energy consumption data; Based on multiple preset energy consumption data intervals, determine the energy consumption data interval to which the predicted energy consumption data belongs to obtain the target energy consumption data interval; Schedule the target energy-saving management plan corresponding to the energy consumption data range, and execute the target energy-saving management plan to achieve the energy-saving management of green buildings; among them, there is a preset energy-saving management plan for each energy consumption data range.

[0054] For example, set a segmentation threshold. The range with energy consumption higher than the threshold is set as the first energy consumption data range, and the range with energy consumption lower than the threshold is set as the second energy consumption data range. When the predicted energy consumption data of a green building is in the second energy consumption data range, no management may be required. When the predicted energy consumption data of a green building is in the first energy consumption data range, the energy-saving management plan preset for the first energy consumption data range can be executed. The target energy-saving management plan can be set as: turning off electrical equipment with a lower priority level, restricting the use of air conditioners, restricting the number of lights used, and so on.

[0055] It should be noted that the above is only an example of the embodiments of the present invention, and more energy consumption data ranges and corresponding energy-saving management plans can also be set to achieve more efficient energy-saving management of green buildings.

[0056] A green building energy-saving design method based on deep learning provided by the present invention constructs a green building energy consumption prediction model by using a deep learning model, and uses the historical sample data as data support. The green building energy consumption prediction model is optimized by using a multi-strategy collaborative optimization algorithm to obtain an optimized green building energy consumption prediction model. Then, the optimized green building energy consumption prediction model can be scheduled to analyze the real-time building energy consumption characteristics to achieve energy consumption prediction. Finally, energy-saving management can be carried out according to the predicted energy consumption data, which can not only effectively improve the energy-saving management efficiency and conduct real-time energy-saving management of green buildings, but also be more conducive to the staff to design energy-saving management.

[0057] As Figure 2 shown, the present invention provides a green building energy-saving design system based on deep learning, including: a historical data acquisition module 21, a deep learning module 22, and a building energy-saving management module 23; The historical data acquisition module 21 is used to collect the historical building energy consumption characteristics corresponding to the target green building and the historical energy consumption data corresponding to the historical building energy consumption characteristics, and after preprocessing the historical building energy consumption characteristics and the historical energy consumption data corresponding to the historical building energy consumption characteristics, obtain historical sample data; The deep learning module 22 is used to construct a green building energy consumption prediction model by using a deep learning model, and use the historical sample data as data support. The green building energy consumption prediction model is optimized by using a multi-strategy collaborative optimization algorithm to obtain an optimized green building energy consumption prediction model; The building energy conservation management module 23 is used to collect the real-time building energy consumption characteristics corresponding to the target green building, schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics, determine the predicted energy consumption data, and perform energy conservation management of the green building according to the predicted energy consumption data.

[0058] The embodiment of the present invention provides a green building energy conservation design system based on deep learning, which can execute the above method technical solution, and its principle and beneficial effects are similar, so they will not be elaborated here.

[0059] After considering the specification and practicing the invention disclosed herein, those skilled in the art will readily conceive of other embodiments of the present invention. The present invention is intended to cover any variations, uses, or adaptations of the present invention, which follow the general principles of the present invention and include common general knowledge or conventional technical means in the technical field not disclosed in the present invention. It should be understood that the present invention is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A green building energy-saving design method based on deep learning, characterized in that, Including: Collecting the historical building energy consumption characteristics corresponding to the target green building and the historical energy consumption data corresponding to the historical building energy consumption characteristics, and obtaining historical sample data after preprocessing the historical building energy consumption characteristics and the historical energy consumption data corresponding to the historical building energy consumption characteristics; Constructing a green building energy consumption prediction model using a deep learning model, and using the historical sample data as data support, optimizing the green building energy consumption prediction model using a multi-strategy collaborative optimization algorithm to obtain an optimized green building energy consumption prediction model; Collecting the real-time building energy consumption characteristics corresponding to the target green building, scheduling the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics, determining the predicted energy consumption data, and performing energy conservation management of the green building according to the predicted energy consumption data.

2. The green building energy-saving design method based on deep learning according to claim 1, characterized in that, Collecting the historical building energy consumption characteristics corresponding to the target green building and the historical energy consumption data corresponding to the historical building energy consumption characteristics, and obtaining historical sample data after preprocessing the historical building energy consumption characteristics and the historical energy consumption data corresponding to the historical building energy consumption characteristics, including: Collecting the average daily temperature, daily maximum temperature, daily minimum temperature, date type, season type, and weather of the target green building on any day to obtain the historical building energy consumption characteristics corresponding to the green building; For the historical building energy consumption characteristics corresponding to any green building, collecting the historical energy consumption data corresponding to the historical building energy consumption characteristics; After normalizing the historical building energy consumption characteristics, obtaining the normalized historical building energy consumption characteristics, constructing the normalized historical building energy consumption characteristics as input samples, and constructing the historical energy consumption data corresponding to the historical building energy consumption characteristics as expected output labels to obtain historical sample data.

3. The green building energy-saving design method based on deep learning according to claim 1, characterized in that, Constructing a green building energy consumption prediction model using a deep learning model, including: constructing a green building energy consumption prediction model using a CNN-LSTM model.

4. The green building energy-saving design method based on deep learning according to any one of claims 1-3, characterized in that, Using the historical sample data as data support, optimizing the green building energy consumption prediction model using a multi-strategy collaborative optimization algorithm to obtain an optimized green building energy consumption prediction model, including: Initializing the hyperparameters of the green building energy consumption prediction model using a chaotic mapping initialization method or a random initialization method, encoding the hyperparameters to obtain an encoded vector, and obtaining multiple different encoded vectors; Using the input samples in the historical sample data as the actual input and the expected output labels in the historical sample data as the expected output to obtain the fitness corresponding to each encoded vector; According to the order of the fitness corresponding to the encoded vectors from large to small, dividing all the encoded vectors into an optimal region search population, a collaborative balance search population, and a fast global search population in turn; For any first target encoded vector in the optimal region search population, searching the first target encoded vector using a boundary probability transfer strategy to obtain the searched first target encoded vector; For any second target encoded vector in the collaborative balance search population, searching the second target encoded vector using a spiral balance search strategy to obtain the searched second target encoded vector; For any third target coding vector in the fast global search population, the optimal attraction and worst warning search strategy is used to search the third target coding vector to obtain the third target coding vector after search; Judge whether the optimization end condition is satisfied. If so, based on the first target coding vector after search, the second target coding vector after search, and the third target coding vector after search, determine the optimized green building energy consumption prediction model. Otherwise, return to the step of determining the fitness corresponding to each coding vector.

5. The green building energy-saving design method based on deep learning according to claim 4, characterized in that For any first target coding vector in the best region search population, the boundary probability transfer strategy is used to search the first target coding vector to obtain the first target coding vector after search, including: For any first target coding vector in the best region search population, based on the boundary of each dimension hyperparameter, obtain the boundary transfer value corresponding to the first target coding vector as: Among them, represents the t -th dimension hyperparameter corresponding to the i-th first target coding vector in the j -th optimization process, represents the i -th dimension hyperparameter corresponding to the j -th first target coding vector corresponding boundary transfer value, where i = 1, 2, …, L, L represents the total number of first target coding vectors, j = 1, 2, …, K, K represents the total dimension of hyperparameters, represents the upper limit of the j-th dimension hyperparameter, represents the lower limit of the j-th dimension hyperparameter, represents the first random number between (0, 1), represents the second random number between (0, 1), represents the j -th dimension hyperparameter of the optimal coding vector; For any first target coding vector in the best region search population, obtain the neighborhood search value of the first target coding vector as: Among them, represents the neighborhood search value of the first target coding vector, represents the j-th dimensional hyperparameter corresponding to a random other first target coding vector, represents pi, represents the preset maximum number of optimizations, and cos represents the cosine function, represents the third random number between (0, 0.55); The first decision factor is generated as: ; where represents the first decision factor, represents the control hyperparameter corresponding to the first decision factor and is set to 0.499; represents the exponential function with the natural constant e as the base; Generate a fourth random number between (0, 1) , and determine the fourth random number whether it is less than the first decision factor , if so, use the neighborhood search value as the first target coding vector after the search, otherwise perform the fitness judgment of the boundary transfer value; Judge whether the fitness value of the boundary transfer value is greater than the fitness corresponding to the original first target coding vector. If so, use the boundary transfer value as the first target coding vector after search. Otherwise, directly use the original first target coding vector as the first target coding vector after search.

6. The green building energy-saving design method based on deep learning according to claim 5, characterized in that For any second target coding vector in the collaborative balance search population, the spiral balance search strategy is used to search the second target coding vector to obtain the second target coding vector after search, including: For any second target coding vector in the collaborative balance search population, obtain the fitness corresponding to the second target coding vector; According to the fitness corresponding to the second target coding vector, arrange the second target coding vectors in descending order to obtain the arranged second target coding vectors; Based on the arranged second target coding vectors, for the m-th second target coding vector, search the second target coding vector to obtain the second target coding vector after search as: Among them, represents the t -th m hyperparameter of the second target coding vector after the j -th permutation during the j -th optimization process, where K = 1, 2, …, K, represents the total dimension of the hyperparameters, m = M / 2 + 1, M / 2 + 2, …, M, where M represents the total number of the second target coding vectors and M is an even number; represents the m -th hyperparameter of the second target coding vector after the j -th search, represents the natural constant function, represents the spiral search control factor, and ; represents the variation coefficient and is set to 5; represents pi, represents the cosine function, represents the exponential function with the natural constant e as the base, represents the preset maximum number of optimization times, represents a random spiral shape control factor uniformly distributed between [-1, 1], and abs represents the absolute value function, represents the j -th hyperparameter of the worst coding vector among all coding vectors; Based on the arranged second target coding vectors, for the g-th second target coding vector, search the second target coding vector to obtain the second target coding vector after search as: Among them, represents the j -th dimensional hyperparameter of the optimal coding vector among all coding vectors, g = 1, 2, …, M / 2, represents the t -th dimensional hyperparameter of the second target coding vector after the g -th permutation in the j -th optimization process, represents the g -th dimensional hyperparameter of the second target coding vector after the j -th search, represents the direction search factor of random dimension -1 or 1.

7. The green building energy-saving design method based on deep learning according to claim 6, characterized in that, For any third target coding vector in the fast global search population, the optimal attraction and worst warning search strategy is used to search the third target coding vector to obtain the third target coding vector after search, including: Randomly generate a fifth random number between (0, 1) ; Determine the fifth random number whether it is less than a preset second decision factor . If so, search for the third target coding vector according to the optimal parameter coding to obtain the third target coding vector after the search. Otherwise, search for the third target coding vector according to the worst parameter coding to obtain the third target coding vector after the search; wherein, the second decision factor is set to t / , where t represents the maximum number of optimization times; According to the optimal parameter coding, search the third target coding vector to obtain the third target coding vector after search as: Among them, represents the t th optimization process of the k th third target coding vector's j dimensional hyperparameter, k = 1, 2, …, P, where P represents the total number of third target coding vectors, j = 1, 2, …, K, K represents the total dimension of hyperparameters, represents the k th third target coding vector's j dimensional hyperparameter after the th search, j represents the th optimal coding vector's dimensional hyperparameter among all coding vectors, represents the sine function, represents the first position weighting factor, represents the second position weighting factor; According to the worst parameter coding, search the third target coding vector to obtain the third target coding vector after search as: Among them, represents the eighth random number between (0, 1), represents the j dimensional hyperparameter of the worst coding vector among all coding vectors, represents the t fitness corresponding to the k th third target coding vector in the th optimization process, represents the constant term and is set to 0.

001.

8. The green building energy-saving design method based on deep learning according to claim 7, characterized in that, Based on the first target coding vector after search, the second target coding vector after search, and the third target coding vector after search, determine the optimized green building energy consumption prediction model, including: Re-obtain the optimal coding vector according to the first target coding vector after searching, the second target coding vector after searching, and the third target coding vector after searching; Use the hyperparameters included in the re-obtained optimal coding vector as the final hyperparameters of the green building energy consumption prediction model to obtain the optimized green building energy consumption prediction model.

9. The green building energy-saving design method based on deep learning according to claim 1, wherein Collect the real-time building energy consumption characteristics corresponding to the target green building, and schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics, determine the predicted energy consumption data, and perform energy conservation management of the green building according to the predicted energy consumption data, including: Collect the real-time building energy consumption characteristics corresponding to the target green building, and schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics, determine the predicted energy consumption data; Based on multiple preset energy consumption data intervals, determine the energy consumption data interval to which the predicted energy consumption data belongs to obtain the target energy consumption data interval; Schedule the target energy conservation management plan corresponding to the target energy consumption data interval and execute the target energy conservation management plan to achieve energy conservation management of the green building; wherein, each energy consumption data interval is preset with an energy conservation management plan.

10. A green building energy-saving design system based on deep learning, characterized in that, Including: A historical data acquisition module, a deep learning module, and a building energy conservation management module; The historical data acquisition module is used to collect the historical building energy consumption characteristics corresponding to the target green building and the historical energy consumption data corresponding to the historical building energy consumption characteristics, and obtain historical sample data after preprocessing the historical building energy consumption characteristics and the historical energy consumption data corresponding to the historical building energy consumption characteristics; The deep learning module is used to construct a green building energy consumption prediction model using a deep learning model, and use the historical sample data as data support to optimize the green building energy consumption prediction model using a multi-strategy collaborative optimization algorithm to obtain the optimized green building energy consumption prediction model; The building energy conservation management module is used to collect the real-time building energy consumption characteristics corresponding to the target green building, and schedule the optimized green building energy consumption prediction model to analyze the real-time building energy consumption characteristics, determine the predicted energy consumption data, and perform energy conservation management of the green building according to the predicted energy consumption data.

Images