Method for predicting flexible regulation and control potential of building body

Abstract

A building body flexible regulation and control potential prediction method comprises the steps that a building group energy consumption model is established based on a back propagation neural network, each energy consumption system sub-model takes characteristic data of influence parameters of energy consumption of each energy consumption system as input, and the maximum power saving rate of the corresponding energy consumption system is taken as a first-level optimization target; outputting energy consumption predicted values of the energy consumption systems in different time periods; determining the energy consumption prediction minimum value of each energy consumption system in different time periods by adopting a distribution entropy improved adaptive grid search particle swarm method and taking the maximum power saving rate of the building group as a second-level optimization target; updating the minimum energy consumption prediction value by using the deviation between the extreme value of each energy consumption system and the power saving rate of the building group; the energy consumption prediction minimum values of all the energy consumption systems in different time periods are spliced according to the time sequence to serve as the energy consumption prediction value of the building group; and taking the proportion of the difference value between the actual energy consumption value of the building group and the predicted energy consumption value of the building group in the actual energy consumption value of the building group as a prediction result of the flexible regulation and control potential of the building body, and realizing accurate prediction.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy operation assessment of high-density building complexes, and in particular, relates to a method for predicting the flexible regulation potential of a building body. Background Art

[0002] Flexible regulation of building energy consumption is of great significance to alleviating the tension of energy use during peak periods, stabilizing building power supply and the initial investment in energy flexible regulation equipment.

[0003] In the prior art, the evaluation of building flexible regulation potential needs to be split and counted based on the electricity consumption data of air conditioning system, lighting office system, power equipment system, etc., and the electricity consumption or energy consumption characteristics of different systems are analyzed from the perspective of time or outdoor climate according to the statistical results to predict the building flexible regulation potential. A large number of statistics on the project will bring greater time cost and investment cost. If the statistical method is inappropriate, the target analysis will be biased. Therefore, there are many methods that use a small number of statistical results to establish a mathematical model, analyze the electricity or energy consumption characteristics of different systems in the building by quantitative means, or optimize the electricity or energy consumption data of the entire building to predict the building's flexible regulation potential. The particle swarm method (GA_PSO) based on the genetic method is used to optimize the generalized regression neural network (GRNN) network method to establish a mathematical model and improve the accuracy of the GRNN model. The high-precision model can better quantify the impact of relevant factors on the building's flexible characteristics, but the model established by the GRNN network method cannot optimize the building's energy consumption and evaluate the building's flexible regulation potential. Existing evaluation technologies focus on the impact of the main factors subjectively identified by quantifiers on electricity or energy consumption, ignoring the impact of the coupling of the main factors on energy consumption characteristics. The influencing factors considered are not comprehensive enough, and it is difficult to objectively obtain the main factors through comparison. In addition, when the modeling method is not appropriate, too many influencing factors can easily lead to large errors in the mathematical model. In addition, the existing technology focuses on integrating different power equipment systems or studying the power or energy characteristics of the building as a whole, predicting the flexible regulation potential of the building itself, and does not optimize the power consumption of different systems and the entire building at the same time, which has limited use value for actual operations. The existing technology proposes orthogonalization experiments and artificial neural network models, and uses standard particle swarm optimization methods to optimize the research objects. The core control parameters of the standard particle swarm optimization method include inertia weight and learning factor, where the learning factor is set to a constant value, and the inertia weight adopts a relatively simple linear decreasing update strategy. This setting method is relatively simple, which may limit the search efficiency of the method, resulting in slow convergence speed and poor solution quality. The adaptive inertia weight adjustment mechanism based on distributed entropy is adopted. This strategy effectively balances the global exploration and local development capabilities of the method, thereby significantly improving the search accuracy and efficiency of the method. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention provides a method for predicting the flexible regulation potential of a building body, which characterizes the flexible regulation potential of a building body from the energy consumption adjustment amount of a building complex, and can obtain the control parameters when the power saving rate of each energy consumption system is the maximum, so as to facilitate reasonable regulation of different energy consumption systems at different times, thereby more accurately predicting the energy consumption of the building complex and realizing efficient prediction of the flexible regulation potential of the building body.

[0005] The present invention adopts the following technical solution.

[0006] The present invention proposes a method for predicting the flexible regulation potential of a building body, wherein the building complex includes a building body and multiple energy consumption systems, including:

[0007] Obtain characteristic data of parameters influencing energy consumption of each energy consumption system;

[0008] A building complex energy consumption model is established based on a back propagation neural network. The building complex energy consumption model includes sub-models of various energy consumption systems. Each energy consumption system sub-model takes the characteristic data of the influencing parameters of the energy consumption of each energy consumption system as input and takes the maximum power saving rate of the corresponding energy consumption system as the first-level optimization target. Each energy consumption system sub-model outputs the predicted energy consumption value of each energy consumption system in different time periods.

[0009] The adaptive grid search particle swarm method improved by distributed entropy is adopted, and the maximum power saving rate of the building group is taken as the second-level optimization goal. The minimum energy consumption forecast value of each energy consumption system in different time periods is determined from the energy consumption forecast values ​​of each energy consumption system in different time periods. When determining the minimum energy consumption forecast value, the deviation between the extreme value of each energy consumption system and the power saving rate of the building group is used to update the minimum energy consumption forecast value.

[0010] The minimum predicted energy consumption values ​​of each energy consumption system in different time periods are spliced ​​together in time sequence as the predicted energy consumption value of the building complex; the proportion of the difference between the actual energy consumption value of the building complex and the predicted energy consumption value of the building complex in the actual energy consumption value of the building complex is used as the prediction result of the flexible regulation potential of the building itself.

[0011] Preferably, obtaining characteristic data of parameters influencing the energy consumption of each energy consumption system includes:

[0012] The historical maximum and minimum values ​​of each influencing parameter are used to construct the historical value range of each influencing parameter;

[0013] The orthogonal experimental method is used to design the simulation scheme of the energy consumption system. Based on the simulation scheme, the maximum and minimum simulation values ​​of each influencing parameter are obtained to construct the simulation value range of each influencing parameter.

[0014] The union of the historical value range and the simulation value range of each influencing parameter is used as the value range of each influencing parameter, and a basic data set is established based on the value ranges of multiple influencing parameters;

[0015] A convolutional neural network is used to extract characteristic data of parameters influencing the energy consumption of each energy consumption system from the basic data set.

[0016] Preferably, the influencing parameters of the energy consumption of each energy consumption system include: indoor temperature, indoor humidity, outdoor temperature, outdoor humidity, indoor wind speed, outdoor wind speed, water temperature, building density, illumination, floor height and power equipment pressure.

[0017] Preferably, in the input layer, the characteristic data of the influencing parameters of the energy consumption of each energy consumption system are normalized;

[0018] Each energy consumption system sub-model also includes: a hidden layer, and the hidden layer of each energy consumption system sub-model adopts a tansig function;

[0019] The output values ​​of each layer of each energy consumption system sub-model are all within the range of [-1,1].

[0020] Preferably, the control parameters of each energy consumption system in different time periods are determined according to the predicted energy consumption values ​​of each energy consumption system in different time periods output by each energy consumption system sub-model, including indoor temperature, indoor humidity, indoor wind speed, water temperature and power equipment pressure.

[0021] Preferably, when the control parameter is not within the value range of the corresponding influencing parameter, the control parameter is added to the basic data set.

[0022] Preferably, an adaptive grid search particle swarm method improved by distributed entropy is adopted, with the maximum power saving rate of the building complex as the second-level optimization goal, and the minimum predicted energy consumption value of each energy consumption system in different time periods is determined from the predicted energy consumption values ​​of each energy consumption system in different time periods, including:

[0023] The energy consumption prediction values ​​of each energy consumption system in different time periods are used as population particles; at the running time t, any population particle Y(t) is in the two population particles Y i (t) and Y j The projection of the distance vector g(t) between (t) forms a set z(t), which satisfies the following relationship:

[0024] z(t)=g(t) T Y(t)

[0025] At running time t, the distance vector g(t) is divided into N intervals according to the population size, and the number of particle projections in the nth interval is counted, denoted as h n (t), where n = 1, 2, ..., N;

[0026] Calculate the population distribution entropy E(t) at running time t, which satisfies the following relationship:

[0027]

[0028] In the formula, s n (t) = h n (t)N;

[0029] Calculate the inertia weight w of the population distribution entropy at runtime t, satisfying the following relationship:

[0030]

[0031] Calculate the first learning factor and the second learning factor to satisfy the following relationship:

[0032]

[0033] In the formula, c 1 is the first learning factor, c 1,ini 、c 1,fin are the first learning factor c 1 The initial and final values ​​of c 2 is the second learning factor, c 2,ini 、c 2,fin are the second learning factor c 1 The initial and final values ​​of K max is the maximum number of iterations, k is the current number of iterations;

[0034] On the basis of the adaptive grid search particle swarm method, based on the first learning factor and the second learning factor, the particle velocity determined by the inertia weight based on the population distribution entropy is linearly updated using the extreme value of the particle position deviation and the extreme value of the population position deviation, and the particle position is linearly updated using the updated particle velocity, satisfying the following relationships respectively:

[0035]

[0036] In the formula, is the velocity of the particle at the kth iteration, w is the inertia weight of the population distribution entropy, c 1 is the first learning factor, c 2 is the second learning factor, r 1 and r 2 are all random numbers distributed between [0,1]. is the extreme value of the particle position at the kth iteration, is the extreme value of the population position at the kth iteration, is the position of the particle at the kth iteration, and t is the running time;

[0037] The optimal value of energy consumption prediction for each energy consumption system is determined according to the obtained particle position, and the power saving rate of the building complex at the kth iteration is calculated. When the error between the power saving rate of the building complex at the kth iteration and the power saving rate of the building complex at the k-1th iteration is not greater than the set threshold, the iteration is terminated, otherwise the first learning factor and the second learning factor are recalculated, and the particle position is updated.

[0038] Preferably, is the extreme value of particle position deviation, representing the extreme value of each energy consumption system; is the extreme value of population position deviation, representing the deviation of the energy saving rate of the building group; Characterize the minimum value of energy consumption prediction;

[0039] On the basis of the adaptive grid search particle swarm method, based on the first learning factor and the second learning factor, the particle velocity determined by the inertia weight based on the population distribution entropy is linearly updated using the extreme values ​​of each energy consumption system and the deviation of the power saving rate of the building complex. The particle position is linearly updated using the updated particle velocity to obtain the updated minimum energy consumption prediction value.

[0040] At the end of the iteration, the minimum energy consumption of each energy consumption system at different times and the control parameters corresponding to the minimum energy consumption are output under the building complex power saving rate.

[0041] Preferably, the threshold is set not to exceed 0.1.

[0042] Preferably, the prediction result of the building body's flexible regulation potential satisfies the following relationship:

[0043]

[0044] In the formula, E potential The flexible control potential of the building itself, E pre is the actual energy consumption of the building complex, E opt is the predicted value of energy consumption of the building complex.

[0045] The beneficial effects of the present invention are that, compared with the prior art, at least the following are included: the basic data set proposed in the present invention is not sample data that affects parameters, but the allowable value range of the data, which enables the basic data set to be adaptively oriented to different building complex operation scenarios; based on the method of linearly decreasing inertia weight, the inertia weight of the population distribution entropy is dynamically updated, which effectively balances the global exploration and local development capabilities of the method, thereby significantly improving the search accuracy and efficiency of the method. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flow chart of a method for predicting the potential of flexible regulation of a building body proposed by the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only embodiments of a part of the present invention, rather than all embodiments. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the protection scope of the present invention.

[0048] The present invention proposes a method for predicting the flexible regulation potential of a building, wherein the building complex includes a building complex and multiple energy consumption systems, and the flexible regulation potential of the building complex is determined according to the energy consumption regulation characteristic quantity of the building complex. In a non-limiting preferred embodiment, the energy consumption system includes but is not limited to an air conditioning system, a water pump system, a heat storage system, and a cold storage system. Figure 1 As shown, the method includes:

[0049] Step 1: Obtain characteristic data of parameters influencing energy consumption of each energy consumption system.

[0050] Specifically, step 1 includes:

[0051] Step 1.1, using the historical maximum value and historical minimum value of each influencing parameter to construct the historical value range of each influencing parameter;

[0052] In a non-limiting preferred embodiment, in order to more comprehensively evaluate the potential of flexible energy regulation of high-density building complexes, the value range of the influencing parameters needs to cover daily parameter changes, and the influencing parameters of energy consumption of each energy consumption system include but are not limited to: indoor temperature, indoor humidity, outdoor temperature, outdoor humidity, indoor wind speed, outdoor wind speed, water temperature, building density, illumination, floor height, and power equipment pressure. The historical maximum and minimum values ​​in the historical data of each influencing parameter are selected to construct the value range of each influencing parameter, ensuring that the value range of the influencing parameter covers various possible situations as much as possible, paving the way for the prediction of the flexible regulation potential of the building body.

[0053] Step 1.2, using the orthogonal experimental method to design a simulation scheme for the energy consumption system, based on the simulation scheme to obtain the maximum and minimum simulation values ​​of each influencing parameter, so as to construct the simulation value range of each influencing parameter;

[0054] In a non-limiting preferred embodiment, an orthogonal experimental method is used to design a basic data set covering multiple ranges of influencing parameters. Those skilled in the art may use other methods to obtain a basic data set covering multiple ranges of influencing parameters.

[0055] In a non-limiting preferred embodiment, each scheme includes at least 11 parameters, and the 11 parameters are arranged and combined according to the traditional method, and the number of simulation schemes involved is very large. The simulation scheme designed based on the orthogonal experimental method can not only cover the influence of the 11 parameters on energy consumption, but also greatly reduce the number of simulation schemes. In this embodiment, the simulation scheme is designed based on the orthogonal experimental method and implemented using SPSS software. The simulation maximum value and the simulation minimum value of the influencing parameter are obtained through each simulation scheme to construct the simulation value range of each influencing parameter.

[0056] Step 1.3, using the union of the historical value range and the simulation value range of each influencing parameter as the value range of each influencing parameter, and establishing a basic data set with the value ranges of multiple influencing parameters;

[0057] The basic data set proposed in the present invention is not sample data that affects parameters, but the permissible value range of the data, which enables the basic data set to be adaptive to different building complex operation scenarios.

[0058] In step 1.4, a convolutional neural network is used to extract characteristic data of the influencing parameters of the energy consumption of each energy consumption system from the basic data set.

[0059] Specifically, a feature extraction model is established based on the convolution kernel of a convolutional neural network, and characteristic data corresponding to the influencing parameters of the energy consumption of each energy consumption system are extracted from the basic data set to reduce the impact of invalid information in the data on the performance of the energy consumption characteristic model, so as to avoid the complexity of the model established by mathematical methods due to many parameters. By reducing the interference of the implicit information entropy of the data corresponding to the influencing parameters, the larger the characteristic value of the influencing parameter, the more effective information is covered.

[0060] Step 2: Establish a building complex energy consumption model based on the back propagation neural network. The building complex energy consumption model includes sub-models of various energy consumption systems. Each energy consumption system sub-model uses the same input layer, so that each energy consumption system sub-model uses the characteristic data of the influencing parameters of the energy consumption of each energy consumption system as input and the maximum power saving rate of the corresponding energy consumption system as the first-level optimization goal. Each energy consumption system sub-model outputs the predicted energy consumption value of each energy consumption system in different time periods.

[0061] The power saving rate of the energy consumption system is a measure of the power saving effect of the energy consumption system under specific conditions. The power saving rate is the ratio of the difference in energy consumption before and after energy saving to the energy consumption before energy saving. The maximum power saving rate of each energy consumption system is taken as the first-level optimization goal to reduce energy consumption locally. However, when each energy consumption system is at the lowest energy consumption, the building complex may not reach the lowest energy consumption. Therefore, it is necessary to further determine the minimum energy consumption forecast value of the building complex based on the energy consumption forecast values ​​of each energy consumption system under different time periods output by each energy consumption system sub-model.

[0062] Specifically, a building complex energy consumption model is established based on a backpropagation neural network (BPNN) to characterize the relationship between each influencing parameter and the energy consumption prediction value of the corresponding energy consumption system. The building complex energy consumption model includes a BPNN sub-model for the air conditioning system, a BPNN sub-model for the water pump system, a BPNN sub-model for the thermal storage system, and a BPNN sub-model for the cold storage system.

[0063] All BPNN sub-models are connected to the same input neuron, so each BPNN sub-model has the same input data, that is, the influencing parameters of each energy consumption system are used as input. Each BPNN sub-model takes the maximum power saving rate of the corresponding energy consumption system as the first-level optimization goal, and each sub-model outputs the energy consumption prediction value of each energy consumption system in different time periods.

[0064] Specifically, each energy consumption system sub-model uses the same input layer, in which characteristic data of the influencing parameters of the energy consumption of each energy consumption system are normalized.

[0065] Each energy consumption system sub-model also includes: hidden layer, output layer; the building complex energy consumption model is as follows:

[0066] (a) Construct the hidden layer using the following relationship:

[0067]

[0068] In the formula, H j is the transfer function of the jth hidden layer, w ij is the weight connecting the i-th input layer to the j-th hidden layer, e i,j is the feature data from the i-th input layer to the j-th hidden layer, b j is the threshold of the hidden layer, n is the number of input layers; s j is the tangent sigmoid function of the jth hidden layer.

[0069] (b) Construct the output layer using the following relationship:

[0070] Y=s j w jk +b k (3)

[0071] Where Y is the output of the output layer, i.e., the predicted value of energy consumption, and w jk is the weight connecting the jth hidden layer to the kth output layer, b k is the threshold of the k-th output layer.

[0072] In the embodiment, the input data is normalized and then input into each BPNN sub-model at the same time. The hidden layer of each BPNN sub-model adopts the tansig function, and the output value of each layer of each BPNN sub-model is within the range of [-1,1], so that the output results of each BPNN sub-model can be fused and spliced ​​in time series.

[0073] In the present invention, the influencing parameters of each energy consumption system are used as the input of each sub-model, and each sub-model is established based on a back-propagation neural network, which ensures the coupling degree between the energy consumption prediction values ​​of each energy consumption system output by each sub-model in different time periods, making the prediction result more accurate and reliable.

[0074] In addition, according to the predicted energy consumption values ​​of each energy consumption system in different time periods output by each energy consumption system sub-model, the control parameters of each energy consumption system in different time periods are determined; among which, the control parameters are influencing parameters that can be adjusted, including indoor temperature, indoor humidity, indoor wind speed, water temperature and power equipment pressure.

[0075] When the control parameter is not within the value range of the corresponding influencing parameter, the control parameter is added to the basic data set.

[0076] Step 3, using the adaptive grid search particle swarm method improved by distributed entropy, taking the maximum power saving rate of the building complex as the second-level optimization goal, and determining the minimum energy consumption forecast value of each energy consumption system in different time periods from the energy consumption forecast values ​​of each energy consumption system in different time periods; wherein, when determining the minimum energy consumption forecast value, the deviation between the extreme values ​​of each energy consumption system and the power saving rate of the building complex is used to update the minimum energy consumption forecast value.

[0077] In step 2, energy consumption has been reduced locally. In order to achieve the lowest energy consumption of the building complex when each energy consumption system is at the lowest energy consumption or close to the lowest energy consumption, that is, to balance global exploration and local development, the present invention proposes an adaptive grid search particle swarm method improved by distributed entropy, with the maximum power saving rate of the building complex as the second-level optimization goal, and the minimum predicted energy consumption of each energy consumption system in different time periods is determined from the predicted energy consumption values ​​of each energy consumption system in different time periods. However, in the process of balancing global exploration and local development, the problem of sacrificing the local should not occur. For example, if any energy consumption system is large or even exceeds the predicted energy consumption value determined in step 2, but at this time, the energy consumption of the building complex is relatively low due to the extremely small energy consumption values ​​of the remaining energy consumption systems. However, at this time, the control parameters of each energy consumption system are not the optimal parameters and need to be excluded. Therefore, self-learning is also required for such local energy consumption extreme value situations to eliminate interference.

[0078] Specifically, step 3 includes:

[0079] Step 3.1: The predicted energy consumption values ​​of each energy consumption system in different time periods are used as population particles; at the running time t, any population particle Y(t) is in the two population particles Y i (t) and Y j The projection of the distance vector g(t) between (t) forms a set z(t), which satisfies the following relationship:

[0080] z(t)=g(t) T Y(t) (4)

[0081] Step 3.2, at running time t, divide the distance vector g(t) into N intervals according to the population size, and count the number of particle projections in the nth interval, denoted as h n (t), where n = 1, 2, ..., N;

[0082] Step 3.3, calculate the population distribution entropy E(t) at the running time t, which satisfies the following relationship:

[0083]

[0084] In the formula, s n (t) = h n (t)N.

[0085] The population distribution entropy is a key indicator to measure the degree of dispersion of the particle swarm in the search space. In the early stage of the adaptive grid search of the particle swarm, the particle swarm is widely scattered and the distribution entropy is large, which promotes the breadth of the global search and helps to find the potential optimal solution area; as the search progresses, the particles gradually gather and the distribution entropy decreases, allowing the method to focus on local fine search and improve local development capabilities. Therefore, by dynamically adjusting the population distribution entropy, the population environment is perceived based on the population distribution entropy, balancing the needs of global exploration and local development.

[0086] Therefore, it is also necessary to further determine the minimum energy consumption forecast value of the building complex based on the energy consumption forecast values ​​of each energy consumption system in different time periods output by each energy consumption system sub-model.

[0087] Step 3.4, calculate the inertia weight w of the population distribution entropy at the running time t, which satisfies the following relationship:

[0088]

[0089] Step 3.5, calculate the first learning factor and the second learning factor, satisfying the following relationship:

[0090]

[0091] In the formula, c 1 is the first learning factor, c 1,ini 、c1,fin are the first learning factor c 1 The initial and final values ​​of c 2 is the second learning factor, c 2,ini 、c 2,fin are the second learning factor c 1 The initial and final values ​​of K max is the maximum number of iterations, and k is the current number of iterations.

[0092] Step 3.6, based on the adaptive grid search particle swarm method, based on the first learning factor and the second learning factor, the particle velocity determined by the inertia weight based on the population distribution entropy is linearly updated using the extreme value of the particle position deviation and the extreme value of the population position deviation, and the particle position is linearly updated using the updated particle velocity, satisfying the following relationships respectively:

[0093]

[0094] In the formula, is the velocity of the particle at the kth iteration, w is the inertia weight of the population distribution entropy, c 1 is the first learning factor, c 2 is the second learning factor, r 1 and r 2 are all random numbers distributed between [0,1]. is the extreme value of the particle position at the kth iteration, is the extreme value of the population position at the kth iteration, is the position of the particle at the kth iteration, and t is the running time;

[0095] In the embodiment, the running time t is set to 1.

[0096] is the extreme value of particle position deviation, representing the extreme value of each energy consumption system; is the extreme value of population position deviation, representing the deviation of the energy saving rate of the building group; Characterize the minimum value of energy consumption prediction; based on the adaptive grid search particle swarm method, based on the first learning factor and the second learning factor, using the extreme values ​​of each energy consumption system and the deviation of the power saving rate of the building group, the particle velocity determined by the inertia weight based on the population distribution entropy is linearly updated, and the particle position is linearly updated using the updated particle velocity to obtain the updated minimum value of energy consumption prediction.

[0097] The adaptive grid search particle swarm method improved based on distributed entropy retains the basic principle of the adaptive grid search particle swarm method. Through the asynchronous update of the first learning factor and the second learning factor in each iteration, the effect of real-time adjustment of the particle velocity update strategy according to the local energy consumption extreme value is achieved, which further enhances the adaptability of the method to complex problems, avoids the problem of sacrificing local problems in the process of balancing global exploration and local development, and finally effectively guides the particle swarm to search for the optimal solution and eliminates the interference of local energy consumption extreme values.

[0098] Step 3.7, determine the optimal value of energy consumption prediction for each energy consumption system based on the particle position obtained in step 3.6, and calculate the power saving rate of the building complex at the kth iteration. When the error between the power saving rate of the building complex at the kth iteration and the power saving rate of the building complex at the k-1th iteration is not greater than the set threshold, end the iteration, otherwise return to step 3.5.

[0099] In the embodiment, the threshold is set not to exceed 0.1.

[0100] In the embodiment, when the power saving rate of the building complex is the highest, through step 3, the control parameters corresponding to the minimum energy consumption and the minimum energy consumption of the air-conditioning system, the water pump system, the heat storage system and the cold storage system at different times are predicted as the regulation parameters of each energy consumption system at different times, thereby improving the convenience of operation.

[0101] Step 4: Concatenate the predicted optimal energy consumption values ​​of each energy consumption system in different time periods in time sequence as the predicted energy consumption value of the building complex; take the proportion of the difference between the actual energy consumption value of the building complex and the predicted energy consumption value of the building complex in the actual energy consumption value of the building complex as the estimation result of the flexible regulation potential of the building entity.

[0102] In the present invention, the energy consumption prediction value of the building complex is characterized by splicing the optimal energy consumption prediction values ​​of the air-conditioning system, the water pump system, the heat storage system and the cold storage system in different time periods in time in a time series, thereby improving the coupling degree between the optimal energy consumption prediction values ​​of each energy consumption system and the energy consumption prediction value of the building complex.

[0103] The prediction results of the building's flexible regulation potential satisfy the following relationship:

[0104]

[0105] In the formula, E potential The flexible control potential of the building itself, E pre is the actual energy consumption of the building complex, E opt is the predicted value of energy consumption of the building complex.

[0106] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0107] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0108] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0109] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting the flexible regulation potential of a building, wherein the building complex includes a building and multiple energy consumption systems, characterized in that: include: Obtain characteristic data of parameters influencing energy consumption of each energy consumption system; A building complex energy consumption model is established based on a back propagation neural network. The building complex energy consumption model includes sub-models of various energy consumption systems. Each energy consumption system sub-model takes the characteristic data of the influencing parameters of the energy consumption of each energy consumption system as input and takes the maximum power saving rate of the corresponding energy consumption system as the first-level optimization target. Each energy consumption system sub-model outputs the predicted energy consumption value of each energy consumption system in different time periods. The adaptive grid search particle swarm method improved by distributed entropy is adopted, and the maximum power saving rate of the building group is taken as the second-level optimization goal. The minimum energy consumption forecast value of each energy consumption system in different time periods is determined from the energy consumption forecast values ​​of each energy consumption system in different time periods. When determining the minimum energy consumption forecast value, the deviation between the extreme value of each energy consumption system and the power saving rate of the building group is used to update the minimum energy consumption forecast value. The minimum predicted energy consumption values ​​of each energy consumption system in different time periods are spliced ​​together in time sequence as the predicted energy consumption value of the building complex; the proportion of the difference between the actual energy consumption value of the building complex and the predicted energy consumption value of the building complex in the actual energy consumption value of the building complex is used as the prediction result of the flexible regulation potential of the building itself.

2. The method for predicting the building's flexibility control potential according to claim 1, characterized in that: Obtain characteristic data of parameters influencing energy consumption of each energy consumption system, including: The historical maximum and minimum values ​​of each influencing parameter are used to construct the historical value range of each influencing parameter; The orthogonal experimental method is used to design the simulation scheme of the energy consumption system. Based on the simulation scheme, the maximum and minimum simulation values ​​of each influencing parameter are obtained to construct the simulation value range of each influencing parameter. The union of the historical value range and the simulation value range of each influencing parameter is used as the value range of each influencing parameter, and a basic data set is established based on the value ranges of multiple influencing parameters; A convolutional neural network is used to extract characteristic data of parameters influencing the energy consumption of each energy consumption system from the basic data set.

3. The method for predicting the building's flexibility control potential according to claim 2, characterized in that: The parameters affecting the energy consumption of each energy consumption system include: indoor temperature, indoor humidity, outdoor temperature, outdoor humidity, indoor wind speed, outdoor wind speed, water temperature, building density, illumination, floor height and power equipment pressure.

4. The method for predicting the building's flexibility control potential according to claim 1, characterized in that: In the input layer, the characteristic data of the parameters affecting the energy consumption of each energy consumption system are normalized; Each energy consumption system sub-model also includes: a hidden layer, and the hidden layer of each energy consumption system sub-model adopts a tansig function; The output values ​​of each layer of each energy consumption system sub-model are all within the range of [-1,1].

5. The method for predicting the building's flexibility control potential according to claim 3 is characterized in that: According to the predicted energy consumption values ​​of each energy consumption system in different time periods output by each energy consumption system sub-model, the control parameters of each energy consumption system in different time periods are determined, including indoor temperature, indoor humidity, indoor wind speed, water temperature and power equipment pressure.

6. The method for predicting the potential of flexible regulation of a building according to claim 5, characterized in that: When the control parameter is not within the value range of the corresponding influencing parameter, the control parameter is added to the basic data set.

7. The method for predicting the building's flexibility control potential according to claim 1, characterized in that: The adaptive grid search particle swarm method improved by distributed entropy is adopted. The maximum power saving rate of the building complex is taken as the second-level optimization goal. The minimum energy consumption forecast value of each energy consumption system in different time periods is determined from the energy consumption forecast values ​​of each energy consumption system in different time periods, including: The energy consumption prediction values ​​of each energy consumption system in different time periods are used as population particles; at the running time t, any population particle Y(t) is in the two population particles Y i (t) and Y j The projection of the distance vector g(t) between (t) forms a set z(t), which satisfies the following relationship: z(t)=g(t) T Y(t) At running time t, the distance vector g(t) is divided into N intervals according to the population size, and the number of particle projections in the nth interval is counted, denoted as h n (t), where n = 1, 2, ..., N; Calculate the population distribution entropy E(t) at running time t, which satisfies the following relationship: In the formula, s n (t) = h n (t)N; Calculate the inertia weight w of the population distribution entropy at runtime t, satisfying the following relationship: Calculate the first learning factor and the second learning factor to satisfy the following relationship: In the formula, c1 is the first learning factor, c 1,ini 、c 1,fin are the initial and final values ​​of the first learning factor c1, c2 is the second learning factor, c 2,ini 、c 2,fin are the initial and final values ​​of the second learning factor c1, K max is the maximum number of iterations, k is the current number of iterations; On the basis of the adaptive grid search particle swarm method, based on the first learning factor and the second learning factor, the particle velocity determined by the inertia weight based on the population distribution entropy is linearly updated using the extreme value of the particle position deviation and the extreme value of the population position deviation, and the particle position is linearly updated using the updated particle velocity, satisfying the following relationships respectively: In the formula, is the velocity of the particle at the kth iteration, w is the inertia weight of the population distribution entropy, c1 is the first learning factor, c2 is the second learning factor, r1 and r2 are both random numbers distributed between [0,1]. is the extreme value of the particle position at the kth iteration, is the extreme value of the population position at the kth iteration, is the position of the particle at the kth iteration, and t is the running time; The optimal value of energy consumption prediction for each energy consumption system is determined according to the obtained particle position, and the power saving rate of the building complex at the kth iteration is calculated. When the error between the power saving rate of the building complex at the kth iteration and the power saving rate of the building complex at the k-1th iteration is not greater than the set threshold, the iteration is terminated, otherwise the first learning factor and the second learning factor are recalculated, and the particle position is updated.

8. The method for predicting the building's flexibility control potential according to claim 7, characterized in that: is the extreme value of particle position deviation, representing the extreme value of each energy consumption system; is the extreme value of population position deviation, representing the deviation of the energy saving rate of the building group; Characterize the minimum value of energy consumption prediction; On the basis of the adaptive grid search particle swarm method, based on the first learning factor and the second learning factor, the particle velocity determined by the inertia weight based on the population distribution entropy is linearly updated using the extreme values ​​of each energy consumption system and the deviation of the power saving rate of the building complex. The particle position is linearly updated using the updated particle velocity to obtain the updated minimum energy consumption prediction value.

9. The method for predicting the building's flexibility control potential according to claim 7, characterized in that: Set the threshold to no more than 0.

1.

10. The method for predicting the building's flexibility control potential according to claim 8, characterized in that: At the end of the iteration, the minimum energy consumption of each energy consumption system at different times and the control parameters corresponding to the minimum energy consumption are output under the building complex power saving rate.

11. The method for predicting the building's flexibility control potential according to claim 1, characterized in that: The prediction results of the building's flexible regulation potential satisfy the following relationship: In the formula, E potential The flexible control potential of the building itself, E pre is the actual energy consumption of the building complex, E opt is the predicted value of energy consumption of the building complex.

Images