Load prediction method and device of combined cooling heating and power system, terminal equipment and storage medium

By frequency decomposing and environmental factor integration of load data of the hot and hot power triple supply system, an accurate load prediction sequence is generated, which solves the problem of failure to fully consider environmental factors and undecomposed historical data in the prior art, and improves load prediction accuracy and system operation efficiency.

CN118944062APending Publication Date: 2024-11-12GUANGDONG POWER GRID CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410974869.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing load prediction method for hot and hot power triple supply system fails to fully consider environmental factors and does not decompose historical load data, resulting in a large deviation from the actual load, which may lead to mismatch between supply and demand and energy waste.

Method used

By frequency decomposing the hot and hot electric load data, a subsequence with different frequency distribution characteristics is generated, and a preset load prediction model is input into the environmental factor data (such as temperature, humidity, wind speed, wind direction) to extract the sequence characteristics and correlation characteristics. , iterative training is performed to generate accurate load prediction sequences.

Benefits of technology

It improves the accuracy of load prediction of the hot and hot power triple supply system, reduces supply and demand mismatch, reduces energy waste, and enhances the adaptability and robustness of the prediction model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118944062A_ABST
    Figure CN118944062A_ABST
Patent Text Reader

Abstract

The invention discloses a load prediction method and device of a combined cooling heating and power system, terminal equipment and a storage medium. Inputting the cold load historical sub-sequence, the heat load historical sub-sequence and the electric load historical sub-sequence with different frequency distribution characteristics and environment factors (such as temperature, humidity, wind speed and wind direction) at the current moment into a preset load prediction model; therefore, the load prediction model can learn and extract internal rules and external influence factors of load change based on the decomposed load subsequences and the environmental factor data, accurately predicts a target subsequence corresponding to the current moment, and performs integration on the target subsequences to obtain a load prediction result. According to the method, the prediction sequences of the cold load, the heat load and the electric load corresponding to the current moment can be generated, so that the load prediction sequence of the combined cooling heating and power system can be obtained more accurately, the situation of mismatching between supply and demand is reduced, and energy waste can be reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of energy load forecasting, and in particular to a load forecasting method, device, terminal equipment and storage medium for a combined cooling, heating and power system. Background Art

[0002] Combined Cooling Heating and Power (CCHP) is a system that integrates multiple forms of energy. It realizes the integrated supply of heating, cooling and power generation according to the characteristics and energy demand of different energy sources. By accurately predicting the cooling load, heating load and electricity load of the CCHP system, the imbalance between supply and demand can be reduced, avoiding energy waste caused by the imbalance between supply and demand.

[0003] However, many existing load forecasting methods currently rely mainly on historical load data to build forecasting models, ignoring the impact of environmental factors on cooling, heating and electricity loads, such as temperature, humidity, wind speed, etc. As the system runs longer and the external environment changes, the load characteristics will also change accordingly. If environmental factors are not included in the training of the forecasting model and the actual forecasting application, there will be a large deviation between the load forecast results and the actual load; and the existing forecasting methods directly input historical load data into the forecasting model without any decomposition of the historical load data. Since the original historical load data often contains complex fluctuations, trends and periodic components, different types of loads (such as cooling load, heating load, electricity load) and load change characteristics on different time scales will be different. Directly inputting the overall data into the model will make the model unable to fully adapt to and capture these differences, resulting in poor prediction performance of the model. The existing load forecasting method for the trigeneration system has the problem of not fully considering environmental factors and not performing any decomposition processing on historical load data. As a result, the sequence of various loads in the trigeneration system cannot be accurately predicted, which may cause a mismatch between supply and demand in the operation of the trigeneration system, thus leading to energy waste. Summary of the invention

[0004] The embodiments of the present invention provide a load prediction method, device, terminal equipment and storage medium for a combined cooling, heating and power (CCHP) system, which can not only perform frequency decomposition on various load sequences of historical CCHP load data before prediction, but also synchronously obtain corresponding environmental factor data as a consideration, thereby more accurately deriving the load prediction sequence of the CCHP system. This can effectively solve the problem that the prior art cannot accurately predict the sequence of various loads of the CCHP system, which may cause a mismatch between supply and demand in the CCHP system during operation, thereby leading to energy waste.

[0005] An embodiment of the present invention provides a load forecasting method for a combined cooling, heating and power system, comprising:

[0006] Obtaining the environmental factors corresponding to the combined cooling, heating and power system of a region at the current moment and the cooling, heating and power load data at several historical moments; wherein the environmental factors include: temperature, humidity, wind speed and wind direction; the cooling, heating and power load data include: cooling load historical sequence, heating load historical sequence and power load historical sequence;

[0007] Decomposing each cooling, heating and electricity load data, generating a number of cooling load history subsequences, heating load history subsequences and electricity load history subsequences with different frequency distribution characteristics;

[0008] Input each cooling load history subsequence, each heating load history subsequence, each electric load history subsequence and environmental factors into a preset load forecasting model, so that the load forecasting model generates a plurality of cooling load target subsequences, heating load target subsequences and electric load target subsequences based on the extracted sequence features; wherein the sequence features are used to characterize the correlation features between the load sequences and the environmental factors;

[0009] Integrate each cooling load target subsequence, each heating load target subsequence and each electric load target subsequence respectively to generate a cooling load prediction sequence, a heating load prediction sequence and an electric load prediction sequence corresponding to the combined cooling, heating and power supply system at the current moment;

[0010] The training process of the preset load forecasting model includes:

[0011] Taking the decomposed cooling load subsequence sample data, the decomposed heating load subsequence sample data, the decomposed electric load subsequence sample data and the environmental factor sample data of each sample data as input, and taking the actual cooling load subsequence, heating load subsequence and electric load subsequence of each sample data as output, the load prediction model is iteratively trained until the model converges to generate a preset load prediction model; wherein the environmental factor sample data includes: sample temperature, sample humidity, sample wind speed and sample wind direction.

[0012] Preferably, the step of decomposing each cooling, heating and electricity load data to generate a plurality of cooling load history subsequences, heating load history subsequences and electricity load history subsequences with different frequency distribution characteristics includes:

[0013] Based on variational modal decomposition and preset modal quantity values, the cooling, heating and electric load data are decomposed into a plurality of first cooling load history subsequences with different frequency distribution characteristics, a plurality of first heating load history subsequences with different frequency distribution characteristics and a plurality of first electric load history subsequences with different frequency distribution characteristics;

[0014] Reconstructing each first cooling load history subsequence according to the sample entropy value corresponding to each first cooling load history subsequence to generate a plurality of cooling load history subsequences with different frequency distribution characteristics; wherein the number of cooling load history subsequences is less than the number of first cooling load history subsequences;

[0015] Reconstructing each first heat load history subsequence according to the sample entropy value corresponding to each first heat load history subsequence to generate a plurality of heat load history subsequences with different frequency distribution characteristics; wherein the number of heat load history subsequences is less than the number of first heat load history subsequences;

[0016] According to the sample entropy values ​​corresponding to each first electric load history subsequence, each first electric load history subsequence is reconstructed to generate a plurality of electric load history subsequences with different frequency distribution characteristics; wherein the number of electric load history subsequences is less than the number of first electric load history subsequences.

[0017] Preferably, the reconstructing each first cooling load historical subsequence according to the sample entropy value corresponding to each first cooling load historical subsequence to generate a plurality of cooling load historical subsequences with different frequency distribution characteristics includes:

[0018] Divide each first cooling load history subsequence according to the sample entropy value to generate a plurality of sequence groups; wherein the difference in the sample entropy value between any two first cooling load history subsequences in each sequence group is not greater than a preset entropy value threshold;

[0019] For each sequence group, each first cooling load historical subsequence in the sequence group is reconstructed to generate a cooling load historical subsequence;

[0020] Among them, different cooling load history subsequences correspond to different frequency distribution characteristics.

[0021] Preferably, the sequence characteristics further include: frequency characteristics;

[0022] The load forecasting model generates a plurality of cooling load target subsequences, heating load target subsequences and electric load target subsequences based on the extracted sequence features, including:

[0023] The load forecasting model extracts frequency features corresponding to each cooling load history subsequence, each heating load history subsequence, and each electric load history subsequence, and simultaneously extracts correlation features for characterizing a load sequence and an environmental factor;

[0024] The load prediction model generates, based on each frequency feature and each associated feature, a plurality of cooling load target subsequences to be output corresponding to each cooling load history subsequence, a plurality of heating load target subsequences to be output corresponding to each heating load history subsequence, and a plurality of electric load target subsequences to be output corresponding to each electric load history subsequence;

[0025] For each cooling load historical subsequence, the corresponding plurality of cooling load target subsequences to be output are optimized and solved, and a cooling load target subsequence is output;

[0026] For each heat load history subsequence, the corresponding several heat load target subsequences to be output are optimized and solved, and a heat load target subsequence is output;

[0027] For each electric load historical subsequence, a number of corresponding electric load target subsequences to be output are optimized and solved, and an electric load target subsequence is output.

[0028] Preferably, for each cooling load historical subsequence, the corresponding plurality of cooling load target subsequences to be output are optimized and solved, and a cooling load target subsequence is output, including:

[0029] For each cooling load historical subsequence, the corresponding several cooling load target subsequences to be output are respectively used as individuals in the population, and any two different individuals in the population are randomly selected as parent individuals;

[0030] Repeat the final individual determination operation until the current number of iterations is the same as the preset number of iterations, and randomly select a target individual as the final individual;

[0031] The cooling load target subsequence to be output corresponding to the final individual is used as the cooling load target subsequence;

[0032] The final individual determination operation includes:

[0033] Get the current iteration number;

[0034] When the current number of iterations is less than the preset number of iterations, a horizontal crossover operation is performed on the two parent individuals based on the same dimensions of the two parent individuals to generate several child individuals;

[0035] For each parent individual, a vertical crossover operation is performed on the parent individual based on different dimensions of the parent individual to generate several offspring individuals;

[0036] Compare the fitness of each offspring individual with the two parent individuals, and mark any two offspring individuals whose fitness is greater than that of the parent individuals as target individuals;

[0037] Adding the current iteration count value to the preset iteration count increment obtains the updated iteration count, and uses the updated iteration count as the iteration count for the next target individual determination operation;

[0038] The two target individuals are used as parent individuals for the next target individual determination operation.

[0039] Preferably, the step of comparing the fitness of each offspring individual with the two parent individuals, and marking any two offspring individuals having fitness greater than that of the parent individuals as target individuals, comprises:

[0040] Based on the competition operator mechanism, the fitness of each offspring individual is compared with that of the two parent individuals, and then the offspring individuals with fitness greater than that of the parent individuals are taken as the individuals to be marked;

[0041] Randomly mark two target individuals from each individual to be marked.

[0042] Preferably, when the load forecasting model is iteratively trained, it also includes:

[0043] During each training, the decomposed cooling load subsequence sample data, the decomposed heating load subsequence sample data, the decomposed electric load subsequence sample data and the environmental factor sample data of each sample data are input into the load forecasting model, so that the load forecasting model generates a sequence forecast result corresponding to each sample data based on the extracted correlation features for characterizing the load sequence and the environmental factor;

[0044] The sequence prediction results corresponding to each sample data are compared with the actual cooling load subsequence, heating load subsequence and electric load subsequence of each sample data, and the network parameters of the load prediction model are adjusted according to the comparison results.

[0045] Based on the above method embodiments, the present invention provides corresponding device embodiments.

[0046] An embodiment of the present invention provides a load prediction device for a combined cooling, heating and power system, comprising: a data acquisition module, a load sequence decomposition module, a load subsequence prediction module and a subsequence integration module;

[0047] The data acquisition module is used to obtain the environmental factors corresponding to the cooling, heating and power trigeneration system of a region at the current moment and the cooling, heating and power load data at several historical moments; wherein the environmental factors include: temperature, humidity, wind speed and wind direction; the cooling, heating and power load data include: cooling load historical sequence, heating load historical sequence and power load historical sequence;

[0048] The load sequence decomposition module is used to decompose each cooling, heating and electricity load data to generate a number of cooling load history subsequences, heating load history subsequences and electricity load history subsequences with different frequency distribution characteristics;

[0049] The load subsequence prediction module is used to input each cooling load history subsequence, each heating load history subsequence, each electric load history subsequence and environmental factors into a preset load prediction model, so that the load prediction model generates a plurality of cooling load target subsequences, heating load target subsequences and electric load target subsequences based on the extracted sequence features; wherein the sequence features are used to characterize the correlation features between the load sequence and the environmental factors;

[0050] The subsequence integration module is used to integrate each cooling load target subsequence, each heating load target subsequence and each electric load target subsequence respectively, and generate a cooling load prediction sequence, a heating load prediction sequence and an electric load prediction sequence corresponding to the combined cooling, heating and power supply system at the current moment;

[0051] The training process of the preset load forecasting model includes:

[0052] Taking the decomposed cooling load subsequence sample data, the decomposed heating load subsequence sample data, the decomposed electric load subsequence sample data and the environmental factor sample data of each sample data as input, and taking the actual cooling load subsequence, heating load subsequence and electric load subsequence of each sample data as output, the load prediction model is iteratively trained until the model converges to generate a preset load prediction model; wherein the environmental factor sample data includes: sample temperature, sample humidity, sample wind speed and sample wind direction.

[0053] Based on the above method embodiments, the present invention provides corresponding terminal device embodiments.

[0054] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the load prediction method for a trigeneration system described in the above-mentioned embodiment of the invention.

[0055] Based on the above method embodiments, the present invention provides corresponding storage medium item embodiments.

[0056] Another embodiment of the present invention provides a storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the load prediction method for a trigeneration system described in the above-mentioned embodiment of the invention.

[0057] The following beneficial effects are achieved by implementing the present invention:

[0058] The embodiment of the present invention provides a load prediction method, device, terminal equipment and storage medium for a combined cooling, heating and power system. The present invention can decompose each type of load data (such as cooling load, heating load and electric load) to generate a plurality of subsequences with different frequency distribution characteristics, and then input each cooling load historical subsequence, heating load historical subsequence, electric load historical subsequence and environmental factors (such as temperature, humidity, wind speed and wind direction) at the current moment with different frequency distribution characteristics into a preset load prediction model, so that the load prediction model can accurately capture the frequency change characteristics of various loads on different time scales based on the decomposed various load sequences, and learn and extract the correlation characteristics between environmental factors and loads, that is, based on the change law of the load itself and the influence of environmental factors on the load, accurately predict the target subsequence corresponding to the current moment, so that by integrating each target subsequence separately, the prediction sequence of the cooling load, heating load and electric load corresponding to the current moment can be generated. Compared with the prior art, the present invention can not only decompose the frequency of various load sequences of historical cooling, heating and electricity load data before making predictions, but also simultaneously obtain the corresponding environmental factor data as a consideration factor, so that the prediction model can learn and extract the internal laws and external influencing factors of load changes based on the decomposed multi-frequency load sub-sequences and environmental factor data, thereby more accurately deriving the load prediction sequence of the trigeneration system, reducing the mismatch between supply and demand and reducing energy waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flow chart of a load prediction method for a combined cooling, heating and power system provided by an embodiment of the present invention.

[0060] Figure 2 It is a sequence diagram of cooling load history subsequences with different frequency distribution characteristics provided by an embodiment of the present invention.

[0061] Figure 3 It is a sequence diagram of cooling load history subsequences after respective sequence groups are reconstructed, provided by an embodiment of the present invention.

[0062] Figure 4 It is a structural diagram of a deep belief network provided by an embodiment of the present invention.

[0063] Figure 5 It is a flow chart of the decomposition and integration process of the cooling, heating and electric load sequence provided by one embodiment of the present invention.

[0064] Figure 6It is a structural schematic diagram of a load prediction device for a combined cooling, heating and power system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] like Figure 1 FIG. 1 is a flow chart of a load prediction method for a combined cooling, heating and power system provided by an embodiment of the present invention. The load prediction method for the combined cooling, heating and power system comprises:

[0067] Step S1: Obtain the environmental factors corresponding to the combined cooling, heating and power system of a region at the current moment and the cooling, heating and power load data at several historical moments; wherein the environmental factors include: temperature, humidity, wind speed and wind direction; the cooling, heating and power load data include: cooling load historical sequence, heating load historical sequence and power load historical sequence;

[0068] Step S2: Decompose each cooling, heating and electricity load data to generate a number of cooling load history subsequences, heating load history subsequences and electricity load history subsequences with different frequency distribution characteristics;

[0069] Step S3: inputting each cooling load history subsequence, each heating load history subsequence, each electric load history subsequence and environmental factors into a preset load forecasting model, so that the load forecasting model generates a plurality of cooling load target subsequences, heating load target subsequences and electric load target subsequences based on the extracted sequence features; wherein the sequence features are used to characterize the correlation features between the load sequences and the environmental factors;

[0070] The training process of the preset load forecasting model includes:

[0071] Taking the decomposed cooling load subsequence sample data, the decomposed heating load subsequence sample data, the decomposed electric load subsequence sample data and the environmental factor sample data of each sample data as input, taking the actual cooling load subsequence, heating load subsequence and electric load subsequence of each sample data as output, the load forecasting model is iteratively trained until the model converges to generate a preset load forecasting model; wherein the environmental factor sample data includes: sample temperature, sample humidity, sample wind speed and sample wind direction;

[0072] Step S4: Integrate each cooling load target subsequence, each heating load target subsequence and each electric load target subsequence respectively to generate a cooling load prediction sequence, a heating load prediction sequence and an electric load prediction sequence corresponding to the trigeneration system at the current moment.

[0073] For step S1, in a preferred embodiment, the present invention can collect the external environmental conditions (such as temperature, humidity, wind speed, wind direction) at the current time point and the actual usage data of the cooling load, heating load and electric load of the trigeneration system over a period of time in the past (i.e., several historical moments).

[0074] Schematically, environmental factors (such as temperature, humidity, etc.) directly affect the load demand of the CCHP system. For example, in the summer when the temperature is high, the cooling load demand will increase significantly; while in the winter when it is cold, the heating load demand will increase. Therefore, by correlating the current environmental conditions with the future load demand and then predicting it, accurate prediction results can be obtained.

[0075] Moreover, by collecting historical load data, we can analyze the trend and periodicity of load changes over time (such as seasonal changes, daily changes, etc.), the correlation between environmental factors and loads, and the correlation between different types of loads. Combining real-time environmental factors with historical load data can make the load forecasting model that considers the influence of multiple factors more comprehensively reflect the actual load situation and accurately predict the current load demand, so as to reasonably arrange the system's production plan and scheduling strategy to improve the overall efficiency and stability of the system.

[0076] For step S2, in a preferred embodiment, the present invention decomposes the sequence by using a variational modal decomposition method, so as to deeply mine the various information carried by the sequence itself, while reducing the modeling workload without causing signal distortion, and can also reference the sample entropy method to reconstruct the decomposed modal components, and input the reconstructed cooling load historical subsequences, heating load historical subsequences, and electric load historical subsequences into a preset load forecasting model.

[0077] Then, each cooling, heating and electricity load data is decomposed to generate a plurality of cooling load history subsequences, heating load history subsequences and electricity load history subsequences with different frequency distribution characteristics, including:

[0078] Based on variational modal decomposition and preset modal quantity values, the cooling, heating and electric load data are decomposed into a plurality of first cooling load history subsequences with different frequency distribution characteristics, a plurality of first heating load history subsequences with different frequency distribution characteristics and a plurality of first electric load history subsequences with different frequency distribution characteristics;

[0079] Reconstructing each first cooling load history subsequence according to the sample entropy value corresponding to each first cooling load history subsequence to generate a plurality of cooling load history subsequences with different frequency distribution characteristics; wherein the number of cooling load history subsequences is less than the number of first cooling load history subsequences;

[0080] Reconstructing each first heat load history subsequence according to the sample entropy value corresponding to each first heat load history subsequence to generate a plurality of heat load history subsequences with different frequency distribution characteristics; wherein the number of heat load history subsequences is less than the number of first heat load history subsequences;

[0081] According to the sample entropy values ​​corresponding to each first electric load history subsequence, each first electric load history subsequence is reconstructed to generate a plurality of electric load history subsequences with different frequency distribution characteristics; wherein the number of electric load history subsequences is less than the number of first electric load history subsequences.

[0082] In an illustrative manner, when reconstructing each first cooling load historical subsequence, the following steps are specifically included:

[0083] Divide each first cooling load history subsequence according to the sample entropy value to generate a plurality of sequence groups; wherein the difference in the sample entropy value between any two first cooling load history subsequences in each sequence group is not greater than a preset entropy value threshold;

[0084] For each sequence group, each first cooling load historical subsequence in the sequence group is reconstructed to generate a cooling load historical subsequence; wherein different cooling load historical subsequences correspond to different frequency distribution characteristics.

[0085] In a preferred embodiment, variational mode decomposition (VMD) in signal processing determines the frequency center and bandwidth of each component by iteratively searching for the optimal solution of the variational model, thereby adaptively achieving frequency domain decomposition of the signal and effective separation of the components.

[0086] VMD can adaptively divide the cooling load history sequence in the frequency domain by iteratively searching for the optimal solution of the variational model. This division is based on the frequency characteristics of the signal itself, so it can generate subsequences with different frequency distribution characteristics. In the VMD process, the bandwidth and center frequency of each mode (i.e., subsequence) are determined by iterative optimization, and each subsequence will revolve around its specific center frequency and have a corresponding bandwidth, thus presenting different frequency distribution characteristics.

[0087] In a preferred embodiment, the number of decomposed modal components K=5, such as Figure 2 as well as Figure 3As shown, Figure 2 is the sequence diagram of the five first cooling load historical subsequences before reconstruction. Figure 3 The figure is a sequence diagram of only three cooling load historical subsequences after reconstruction. Indicatively, if the five modal components are directly modeled and predicted, the workload will increase significantly. In order to reduce the workload and extract the characteristics of the cooling, heating and electric load sequences, the decomposed modal components are reconstructed using sample entropy to ensure that the extracted sub-signal sequence will not be distorted while reducing the workload.

[0088] Illustratively, in a preferred embodiment, the sample entropy parameters m=2, r=0.2*std are set, and the calculation results of the sample entropy of the cold, hot and electric load sequence are shown in the following table.

[0089] Table 1 Sample entropy of cooling load modal components when K = 5

[0090]

[0091] It can be seen from Table 1 that the sample entropy values ​​of each modal component of the cooling load sequence are different, among which the modal component U2, modal component U3 and modal component U4 are relatively close (that is, the difference in sample entropy values ​​is not greater than the preset entropy value threshold). In order to reduce the workload without affecting the distortion of the subsequence signal, the modal component U2, modal component U3 and modal component U4 can be reconstructed to form a new modal component, defined as modal component U2; thus, a new sequence can be obtained after reconstruction, including: U1, U2 and U3 (originally U5).

[0092] It can be understood that when reconstructing each first heat load historical subsequence and reconstructing each first electric load historical subsequence, the method adopted is the same as that of the cold load. After the variational modal decomposition, the cold, hot and electric loads obtain multiple modal components. In order to reduce the workload of constructing a load prediction model for each modal component, the sample entropy method is used to calculate each modal component and arrange them in descending order. Then, the sample entropy values ​​of each modal component are compared and analyzed, and the modal components corresponding to similar sample entropy values ​​are reconstructed. Schematically, the sample entropy reflects the complexity and irregularity of the time series. By comparing the sample entropy values ​​to reconstruct the modal components, it can be ensured that the reconstructed sequence will not be distorted. The reconstruction method is generally the accumulation method, and the reconstructed modal components are obtained.

[0093] For step S3, in a preferred embodiment, the load forecasting model of the present invention is a deep belief network (DBN), which is a deep neural network that combines unsupervised learning and supervised fine-tuning, and is composed of a plurality of restricted Boltzmann machines (RBM) stacks and a top-level BP neural network, and has a structure as follows: Figure 4 As shown in Figure 1, the layer structure of DBN is shown, including multiple RBM layers and the top BP layer, as well as the connection relationship between them.

[0094] The training process of the preset load forecasting model includes:

[0095] Taking the decomposed cooling load subsequence sample data, the decomposed heating load subsequence sample data, the decomposed electric load subsequence sample data and the environmental factor sample data of each sample data as input, taking the actual cooling load subsequence, heating load subsequence and electric load subsequence of each sample data as output, the load forecasting model is iteratively trained until the model converges to generate a preset load forecasting model; wherein the environmental factor sample data includes: sample temperature, sample humidity, sample wind speed and sample wind direction;

[0096] During each training, the decomposed cooling load subsequence sample data, the decomposed heating load subsequence sample data, the decomposed electric load subsequence sample data and the environmental factor sample data of each sample data are input into the load forecasting model, so that the load forecasting model generates a sequence forecast result corresponding to each sample data based on the extracted correlation features for characterizing the load sequence and the environmental factor;

[0097] The sequence prediction results corresponding to each sample data are compared with the actual cooling load subsequence, heating load subsequence and electric load subsequence of each sample data, and the network parameters of the load prediction model are adjusted according to the comparison results. In schematic form, the prediction results can be compared with the actual values, and the errors (such as mean square error, root mean square error, etc.) can be calculated. According to the error size, the network parameters of the model (including weights and biases) can be adjusted through the back propagation algorithm to reduce the prediction error of the model.

[0098] In a preferred embodiment, the sequence characteristics further include: frequency characteristics;

[0099] The load forecasting model generates a number of cooling load target subsequences, heating load target subsequences and electric load target subsequences based on the extracted sequence features, specifically including:

[0100] The load forecasting model extracts frequency features corresponding to each cooling load history subsequence, each heating load history subsequence, and each electric load history subsequence, and simultaneously extracts correlation features for characterizing a load sequence and an environmental factor;

[0101] The load prediction model generates, based on each frequency feature and each associated feature, a plurality of cooling load target subsequences to be output corresponding to each cooling load history subsequence, a plurality of heating load target subsequences to be output corresponding to each heating load history subsequence, and a plurality of electric load target subsequences to be output corresponding to each electric load history subsequence;

[0102] For each cooling load historical subsequence, the corresponding plurality of cooling load target subsequences to be output are optimized and solved, and a cooling load target subsequence is output;

[0103] For each heat load history subsequence, the corresponding several heat load target subsequences to be output are optimized and solved, and a heat load target subsequence is output;

[0104] For each electric load historical subsequence, a number of corresponding electric load target subsequences to be output are optimized and solved, and an electric load target subsequence is output.

[0105] Among them, when searching for the optimal solution, the present invention adopts the vertical and horizontal optimization algorithm to optimize the deep belief network, so as to obtain the optimal solution, that is, it can output the optimal cold load target subsequence, heat load target subsequence and electric load target subsequence.

[0106] In a preferred embodiment, for each cooling load history subsequence, a number of corresponding cooling load target subsequences to be output are optimized and solved, and a cooling load target subsequence is output, including:

[0107] For each cooling load historical subsequence, the corresponding several cooling load target subsequences to be output are respectively used as individuals in the population, and any two different individuals in the population are randomly selected as parent individuals;

[0108] Repeat the final individual determination operation until the current number of iterations is the same as the preset number of iterations, and randomly select a target individual as the final individual;

[0109] The cooling load target subsequence to be output corresponding to the final individual is used as the cooling load target subsequence;

[0110] The final individual determination operation includes:

[0111] Get the current iteration number;

[0112] When the current number of iterations is less than the preset number of iterations, a horizontal crossover operation is performed on the two parent individuals based on the same dimensions of the two parent individuals to generate several child individuals;

[0113] For each parent individual, a vertical crossover operation is performed on the parent individual based on different dimensions of the parent individual to generate several offspring individuals;

[0114] Compare the fitness of each offspring individual with the two parent individuals, and mark any two offspring individuals whose fitness is greater than that of the parent individuals as target individuals; specifically, the fitness of each offspring individual can be compared with the two parent individuals based on the competition operator mechanism, and then the offspring individuals whose fitness is greater than that of the parent individuals are taken as the individuals to be marked; randomly mark two target individuals from each individual to be marked;

[0115] Adding the current iteration count value to the preset iteration count increment obtains the updated iteration count, and uses the updated iteration count as the iteration count for the next target individual determination operation;

[0116] The two target individuals are used as parent individuals for the next target individual determination operation.

[0117] Illustratively, in the process of optimizing the output heat load target subsequence and the electric load target subsequence, the above-mentioned hybrid operation of horizontal crossover and vertical crossover is used for optimization. Through the hybrid operation of horizontal crossover and vertical crossover, possible solutions can be explored more extensively in the solution space, and the information of the existing optimal solution can be effectively utilized, thereby increasing the possibility of finding the global optimal solution or a solution close to the global optimal solution, thereby improving the accuracy of load forecasting.

[0118] In a preferred embodiment, the deep belief network is optimized using a vertical and horizontal optimization algorithm, including the following steps:

[0119] (1) Initialize the population;

[0120] (2) Calculate the horizontal crossover and compare the competing operators:

[0121] Before the horizontal crossover is performed, all individual particles in the population need to be randomly paired, and any paired particles should not be repeated. Then, all-dimensional crossover calculations are performed on the paired particles. Particles X(i) and X(j) can perform d-dimensional crossover. The calculation formula is as follows:

[0122] M hc (i, d)=r1X(i,d)+(1-r1)X(j,d)+c1(X(i,d)-X(j,d));

[0123] M hc(j, d)=r2X(j,d)+(1-r2)X(i,d)+c2(X(j,d)-X(i,d));

[0124] Wherein: r1 and r1 are random numbers in the range of [-1, 1]; c2 and c1 are random numbers in the range of [-1, 1]; X(i, d) and X(j, d) are the d-th dimensions of particles X(i) and X(j), respectively; Mhc(i, d) and Mhc(j, d) are the d-th dimensional offspring of X(i, d) and X(j, d) after horizontal crossover, respectively.

[0125] (3) Calculate the vertical crossover and compare the competing operators:

[0126] Different from the horizontal crossing method, the vertical crossing is to cross the dimensions of the particle. The calculation formula for the vertical crossing between the mth dimension and the nth dimension of particle X(i) is as follows:

[0127] M vc (i,d m )=rX(i,d m )+(1-r)X(i,d n );

[0128] Where r is a random number in the range of [0, 1]; X(i, dm) and X(i, dn) are the m-th and n-th dimensions of particle X(i) respectively; Mvc(i, dm) is the daughter particle generated by the vertical crossover of the m-th and n-th dimensions of particle X(i); m and n are the dimensions of the particle and are not equal to each other.

[0129] Therefore, the vertical and horizontal optimization algorithm can avoid falling into the local optimal solution to a certain extent through a mixed crossover strategy. Horizontal crossover helps to explore different values ​​in the same dimension, while vertical crossover helps to make combinatorial innovations in different dimensions, thereby escaping the trap of local optimality.

[0130] During the training and application process, the present invention not only focuses on the fitting degree of the current data, but also screens out better solutions through fitness comparison. Therefore, the trained model often has better generalization ability, that is, it can still maintain a high prediction accuracy on new and unseen data.

[0131] In summary, the present invention uses the vertical and horizontal optimization algorithm to optimize the deep belief network, which can significantly improve the accuracy, robustness, convergence speed and generalization ability of load forecasting, and provide more reliable and accurate data support for subsequent energy management, scheduling and planning.

[0132] For step S4, in a preferred embodiment, Figure 5As shown in the flowchart of the decomposition and integration process of the cold, heat and electric load sequences, the present invention can decompose each type of load data (such as cold load, heat load and electric load) to generate a number of subsequences with different frequency distribution characteristics, and then input the cold load historical subsequences, heat load historical subsequences, electric load historical subsequences and the environmental factors at the current moment (such as temperature, humidity, wind speed and wind direction) with different frequency distribution characteristics into a preset load prediction model, so that the load prediction model can accurately predict the target subsequence corresponding to the current moment based on the change law of the load itself and the influence of environmental factors on the load, so that by integrating each target subsequence separately, the prediction sequence of the cold load, heat load and electric load corresponding to the current moment can be generated.

[0133] In a preferred embodiment, the merged sequence may fluctuate due to discontinuity or prediction errors between subsequences. In this case, the embodiment of the present invention may use smoothing techniques (such as moving average, exponential smoothing, etc.) to reduce such fluctuations and make the predicted sequence smoother.

[0134] The present invention decomposes the load data into subsequences with different frequency distribution characteristics, and can perform more refined modeling and prediction for the characteristics of each subsequence, so that the model can capture the complex patterns in the load data, thereby improving the overall prediction accuracy. Different subsequences may be affected by different environmental factors. By predicting these subsequences separately, we can better understand how environmental factors affect load changes, and adjust the prediction model accordingly to make it more adaptable and robust.

[0135] The decomposed subsequences can support multi-scale predictions from short-term to long-term. The present invention can obtain a more comprehensive load prediction result by combining prediction results of different scales. Through accurate load prediction and optimized scheduling, unnecessary energy production and consumption can be reduced, emissions can be reduced, and the goal of energy conservation and emission reduction can be achieved.

[0136] like Figure 6 As shown, based on the above-mentioned various embodiments of the load prediction method for the combined cooling, heating and power trigeneration system, the present invention provides a corresponding device embodiment;

[0137] An embodiment of the present invention provides a load prediction device for a combined cooling, heating and power system, comprising: a data acquisition module, a load sequence decomposition module, a load subsequence prediction module and a subsequence integration module;

[0138] The data acquisition module is used to obtain the environmental factors corresponding to the cooling, heating and power trigeneration system of a region at the current moment and the cooling, heating and power load data at several historical moments; wherein the environmental factors include: temperature, humidity, wind speed and wind direction; the cooling, heating and power load data include: cooling load historical sequence, heating load historical sequence and power load historical sequence;

[0139] The load sequence decomposition module is used to decompose each cooling, heating and electricity load data to generate a number of cooling load history subsequences, heating load history subsequences and electricity load history subsequences with different frequency distribution characteristics;

[0140] The load subsequence prediction module is used to input each cooling load history subsequence, each heating load history subsequence, each electric load history subsequence and environmental factors into a preset load prediction model, so that the load prediction model generates a plurality of cooling load target subsequences, heating load target subsequences and electric load target subsequences based on the extracted sequence features; wherein the sequence features are used to characterize the correlation features between the load sequence and the environmental factors;

[0141] The subsequence integration module is used to integrate each cooling load target subsequence, each heating load target subsequence and each electric load target subsequence respectively, and generate a cooling load prediction sequence, a heating load prediction sequence and an electric load prediction sequence corresponding to the combined cooling, heating and power supply system at the current moment;

[0142] The training process of the preset load forecasting model includes:

[0143] Taking the decomposed cooling load subsequence sample data, the decomposed heating load subsequence sample data, the decomposed electric load subsequence sample data and the environmental factor sample data of each sample data as input, and taking the actual cooling load subsequence, heating load subsequence and electric load subsequence of each sample data as output, the load prediction model is iteratively trained until the model converges to generate a preset load prediction model; wherein the environmental factor sample data includes: sample temperature, sample humidity, sample wind speed and sample wind direction.

[0144] It should be noted that the device embodiments described above are merely schematic, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without paying creative labor.

[0145] Those skilled in the art can clearly understand that, for the sake of convenience and simplicity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0146] Based on the above-mentioned various embodiments of the load prediction method for the trigeneration system, the present invention provides corresponding embodiments of terminal equipment items.

[0147] An embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, a load prediction method for a trigeneration system described in any method embodiment of the present invention is implemented.

[0148] The terminal device may be a computing terminal device such as a desktop computer, a notebook, a palm computer, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0149] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0150] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device or other volatile solid-state storage device.

[0151] Based on the above-mentioned various embodiments of the load prediction method for the trigeneration system, the present invention provides a corresponding storage medium item embodiment.

[0152] An embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a load prediction method for a combined heat, cooling and power trigeneration system as described in any method embodiment of the present invention.

[0153] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each method embodiment described above can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0154] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A load forecasting method for a combined cooling, heating and power system. It is characterized in that include: Obtaining the environmental factors corresponding to the combined cooling, heating and power system of a region at the current moment and the cooling, heating and power load data at several historical moments; wherein the environmental factors include: temperature, humidity, wind speed and wind direction; the cooling, heating and power load data include: cooling load historical sequence, heating load historical sequence and power load historical sequence; Decomposing each cooling, heating and electricity load data, generating a number of cooling load history subsequences, heating load history subsequences and electricity load history subsequences with different frequency distribution characteristics; Input each cooling load history subsequence, each heating load history subsequence, each electric load history subsequence and environmental factors into a preset load forecasting model, so that the load forecasting model generates a plurality of cooling load target subsequences, heating load target subsequences and electric load target subsequences based on the extracted sequence features; wherein the sequence features are used to characterize the correlation features between the load sequences and the environmental factors; Integrate each cooling load target subsequence, each heating load target subsequence and each electric load target subsequence respectively to generate a cooling load prediction sequence, a heating load prediction sequence and an electric load prediction sequence corresponding to the combined cooling, heating and power supply system at the current moment; The training process of the preset load forecasting model includes: Taking the decomposed cooling load subsequence sample data, the decomposed heating load subsequence sample data, the decomposed electric load subsequence sample data and the environmental factor sample data of each sample data as input, and taking the actual cooling load subsequence, heating load subsequence and electric load subsequence of each sample data as output, the load prediction model is iteratively trained until the model converges to generate a preset load prediction model; wherein the environmental factor sample data includes: sample temperature, sample humidity, sample wind speed and sample wind direction.

2. A load prediction method for a combined cooling, heating and power system according to claim 1, It is characterized in that The step of decomposing each cooling, heating and electricity load data to generate a plurality of cooling load history subsequences, heating load history subsequences and electricity load history subsequences with different frequency distribution characteristics includes: Based on variational modal decomposition and preset modal quantity values, the cooling, heating and electric load data are decomposed into a plurality of first cooling load history subsequences with different frequency distribution characteristics, a plurality of first heating load history subsequences with different frequency distribution characteristics and a plurality of first electric load history subsequences with different frequency distribution characteristics; Reconstructing each first cooling load history subsequence according to the sample entropy value corresponding to each first cooling load history subsequence to generate a plurality of cooling load history subsequences with different frequency distribution characteristics; wherein the number of cooling load history subsequences is less than the number of first cooling load history subsequences; Reconstructing each first heat load history subsequence according to the sample entropy value corresponding to each first heat load history subsequence to generate a plurality of heat load history subsequences with different frequency distribution characteristics; wherein the number of heat load history subsequences is less than the number of first heat load history subsequences; According to the sample entropy values ​​corresponding to each first electric load history subsequence, each first electric load history subsequence is reconstructed to generate a plurality of electric load history subsequences with different frequency distribution characteristics; wherein the number of electric load history subsequences is less than the number of first electric load history subsequences.

3. A method for predicting load of a combined cooling, heating and power system according to claim 2, It is characterized in that The reconstructing each first cooling load history subsequence according to the sample entropy value corresponding to each first cooling load history subsequence to generate a plurality of cooling load history subsequences with different frequency distribution characteristics includes: Divide each first cooling load history subsequence according to the sample entropy value to generate a plurality of sequence groups; wherein the difference in the sample entropy value between any two first cooling load history subsequences in each sequence group is not greater than a preset entropy value threshold; For each sequence group, each first cooling load historical subsequence in the sequence group is reconstructed to generate a cooling load historical subsequence; Among them, different cooling load history subsequences correspond to different frequency distribution characteristics.

4. A method for predicting load of a combined cooling, heating and power system according to claim 2, It is characterized in that The sequence characteristics also include: frequency characteristics; The load forecasting model generates a plurality of cooling load target subsequences, heating load target subsequences and electric load target subsequences based on the extracted sequence features, including: The load forecasting model extracts frequency features corresponding to each cooling load history subsequence, each heating load history subsequence, and each electric load history subsequence, and simultaneously extracts correlation features for characterizing a load sequence and an environmental factor; The load prediction model generates, based on each frequency feature and each associated feature, a plurality of cooling load target subsequences to be output corresponding to each cooling load history subsequence, a plurality of heating load target subsequences to be output corresponding to each heating load history subsequence, and a plurality of electric load target subsequences to be output corresponding to each electric load history subsequence; For each cooling load historical subsequence, the corresponding plurality of cooling load target subsequences to be output are optimized and solved, and a cooling load target subsequence is output; For each heat load history subsequence, the corresponding several heat load target subsequences to be output are optimized and solved, and a heat load target subsequence is output; For each electric load historical subsequence, a number of corresponding electric load target subsequences to be output are optimized and solved, and an electric load target subsequence is output.

5. A method for predicting load of a combined cooling, heating and power system according to claim 4, It is characterized in that For each cooling load history subsequence, the corresponding plurality of cooling load target subsequences to be output are optimized and solved, and a cooling load target subsequence is output, including: For each cooling load historical subsequence, the corresponding several cooling load target subsequences to be output are respectively used as individuals in the population, and any two different individuals in the population are randomly selected as parent individuals; Repeat the final individual determination operation until the current number of iterations is the same as the preset number of iterations, and randomly select a target individual as the final individual; The cooling load target subsequence to be output corresponding to the final individual is used as the cooling load target subsequence; The final individual determination operation includes: Get the current iteration number; When the current number of iterations is less than the preset number of iterations, a horizontal crossover operation is performed on the two parent individuals based on the same dimensions of the two parent individuals to generate several child individuals; For each parent individual, a vertical crossover operation is performed on the parent individual based on different dimensions of the parent individual to generate several offspring individuals; Compare the fitness of each offspring individual with the two parent individuals, and mark any two offspring individuals whose fitness is greater than that of the parent individuals as target individuals; Adding the current iteration count value to the preset iteration count increment obtains the updated iteration count, and uses the updated iteration count as the iteration count for the next target individual determination operation; The two target individuals are used as parent individuals for the next target individual determination operation.

6. A method for predicting load of a combined cooling, heating and power system according to claim 5, It is characterized in that The step of comparing the fitness of each offspring individual with the two parent individuals, and marking any two offspring individuals whose fitness is greater than that of the parent individuals as target individuals, includes: Based on the competition operator mechanism, the fitness of each offspring individual is compared with that of the two parent individuals, and then the offspring individuals with fitness greater than that of the parent individuals are taken as the individuals to be marked; Randomly mark two target individuals from each individual to be marked.

7. A method for load prediction of a combined cooling, heating and power system according to claim 1, It is characterized in that When iteratively training the load forecasting model, it also includes: During each training, the decomposed cooling load subsequence sample data, the decomposed heating load subsequence sample data, the decomposed electric load subsequence sample data and the environmental factor sample data of each sample data are input into the load forecasting model, so that the load forecasting model generates a sequence forecast result corresponding to each sample data based on the extracted correlation features for characterizing the load sequence and the environmental factor; The sequence prediction results corresponding to each sample data are compared with the actual cooling load subsequence, heating load subsequence and electric load subsequence of each sample data, and the network parameters of the load prediction model are adjusted according to the comparison results.

8. A load prediction device for a combined cooling, heating and power system. It is characterized in that include: Data acquisition module, load sequence decomposition module, load subsequence prediction module and subsequence integration module; The data acquisition module is used to obtain the environmental factors corresponding to the cooling, heating and power trigeneration system of a region at the current moment and the cooling, heating and power load data at several historical moments; wherein the environmental factors include: temperature, humidity, wind speed and wind direction; the cooling, heating and power load data include: cooling load historical sequence, heating load historical sequence and power load historical sequence; The load sequence decomposition module is used to decompose each cooling, heating and electricity load data to generate a number of cooling load history subsequences, heating load history subsequences and electricity load history subsequences with different frequency distribution characteristics; The load subsequence prediction module is used to input each cooling load history subsequence, each heating load history subsequence, each electric load history subsequence and environmental factors into a preset load prediction model, so that the load prediction model generates a plurality of cooling load target subsequences, heating load target subsequences and electric load target subsequences based on the extracted sequence features; wherein the sequence features are used to characterize the correlation features between the load sequence and the environmental factors; The subsequence integration module is used to integrate each cooling load target subsequence, each heating load target subsequence and each electric load target subsequence respectively, and generate a cooling load prediction sequence, a heating load prediction sequence and an electric load prediction sequence corresponding to the combined cooling, heating and power supply system at the current moment; The training process of the preset load forecasting model includes: Taking the decomposed cooling load subsequence sample data, the decomposed heating load subsequence sample data, the decomposed electric load subsequence sample data and the environmental factor sample data of each sample data as input, and taking the actual cooling load subsequence, heating load subsequence and electric load subsequence of each sample data as output, the load prediction model is iteratively trained until the model converges to generate a preset load prediction model; wherein the environmental factor sample data includes: sample temperature, sample humidity, sample wind speed and sample wind direction.

9. A terminal device, It is characterized in that The invention comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a load prediction method for a combined cooling, heating and power system as claimed in any one of claims 1 to 7 is implemented.

10. A storage medium, It is characterized in that The storage medium includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the load prediction method for a combined heating, cooling and power trigeneration system according to any one of claims 1 to 7.