Building cooling and heating load prediction method based on generative adversarial network

Abstract

The invention provides a building cooling and heating load prediction method based on a generative adversarial network. The method comprises the following steps: constructing a cooling and heating load prediction model and an indoor temperature judgment model; the indoor temperature judgment model can obtain an indoor predicted temperature by taking a mixture of cooling and heating load real data and cooling and heating load prediction data output by the cooling and heating load prediction model as input; based on a generative adversarial network, alternately training the two models to obtain a target cooling and heating load prediction model; the indoor temperature judgment model judges whether the input cooling and heating load data is true or false by judging whether the output indoor predicted temperature and the indoor true temperature are true or false, and then model parameters are optimized; and predicting the cooling and heating load of the building according to the target cooling and heating load prediction model to obtain a cooling and heating load prediction value of the building. According to the invention, the accuracy and convenience of building cooling and heating load prediction are improved, and continuous evolution and updating can be realized through data accumulation.

Description

Technical Field

[0001] The invention relates to the field of dynamic regulation of building air-conditioning systems, and in particular to a method for predicting building cooling and heating loads based on a generative adversarial network. Background Art

[0002] In the field of building air conditioning system control, passive control methods based on proportional-integral-differential (PID) are currently mostly used, that is, the future control parameters are determined based on the parameter feedback of the current system state. Although this method can meet the building's indoor temperature control needs within a certain range, due to the delay and hysteresis of heat transfer and flow in the building air conditioning system, the matching of cold and hot supply is reduced while increasing the system's operating energy consumption. If a predictive control method is adopted, and the operating parameters of the air conditioning system are optimized in advance by predicting the future load demand of the system, the matching of the air conditioning cold and hot supply and the building's heat demand can be improved, and accurate cold and hot supply can be achieved, reducing operating energy consumption.

[0003] High-precision prediction of building cooling and heating loads is a key step in achieving predictive control. Cooling and heating loads are important bases for the operation and regulation of air-conditioning systems. Traditional load calculations are determined based on parameters such as outdoor meteorological parameters, envelope performance, and internal heat sources of buildings. Existing load prediction methods are usually based on physical simulation and data-driven methods. The specific design parameters of the building and external environmental parameters are input through simulation software to simulate the building energy consumption to obtain air-conditioning load data, and the historical data containing meteorological parameters and load parameters are used to establish a data-driven prediction model. However, the existing methods are complex to model and have high requirements on the quantity and quality of historical data. However, there are large differences between actual meteorological conditions and typical meteorological data used in load simulation. At the same time, it is usually difficult to provide complete historical data in operation control, and it is difficult to build an accurate building load prediction model to achieve high-precision predictive control of air-conditioning systems. Summary of the invention

[0004] In view of this, the present invention aims to propose a building cooling and heating load prediction method based on a generative adversarial network to solve the problem of difficulty in achieving accurate real-time load prediction in the actual prediction and control of the air-conditioning system due to the lack of data, and to improve the accuracy and convenience of predictive control.

[0005] To achieve the above object, the technical solution created by the present invention is implemented as follows: A building cooling and heating load prediction method based on a generative adversarial network comprises the following steps: Constructing a cold and hot load prediction model and an indoor temperature judgment model; the indoor temperature judgment model can use a mixture of cold and hot load real data and cold and hot load prediction data output by the cold and hot load prediction model as input to obtain an indoor predicted temperature; Based on the generative adversarial network, the cooling and heating load prediction model and the indoor temperature judgment model are alternately trained to obtain a target cooling and heating load prediction model; the indoor temperature judgment model judges the authenticity of the input cooling and heating load data by judging the authenticity of the indoor predicted temperature output by it and the indoor real temperature, thereby optimizing its model parameters; The cooling and heating loads of the building are predicted according to the target cooling and heating load prediction model to obtain the cooling and heating load prediction value of the building.

[0006] Furthermore, the method for constructing the cooling and heating load prediction model is: A one-dimensional convolutional neural network is used to calculate the cooling and heating load values ​​based on the air-conditioning side parameters to build a cooling and heating load prediction model; the air-conditioning side parameters include supply water temperature, return water temperature, and supply water return water flow rate; The cooling and heating load prediction model is obtained by training using historical data of meteorological parameters as input and predicted values ​​of cooling and heating loads as output; the meteorological parameters include outdoor temperature and humidity, solar radiation intensity, and wind speed.

[0007] Furthermore, the method for constructing the indoor temperature judgment model is: Use one-dimensional convolutional neural network to build indoor temperature judgment model; The indoor temperature judgment model is obtained by training with historical data of cooling and heating loads and historical data of meteorological parameters as input and predicted indoor temperature as output.

[0008] Furthermore, the alternating training method of the cooling and heating load prediction model and the indoor temperature judgment model is: Training the indoor temperature judgment model: The cooling and heating load prediction model is fixed, the real cooling and heating load data and the cooling and heating load prediction data output by the cooling and heating load prediction model are mixed as input, and the indoor temperature prediction model is used to obtain the indoor temperature; Compare the predicted indoor temperature with the actual indoor temperature, and judge the authenticity of the predicted indoor temperature and the actual indoor temperature, as well as the authenticity of the actual data of the cooling and heating loads and the predicted data of the cooling and heating loads; Calculate the loss function and update the model parameters based on the judgment results; Training the cooling and heating load prediction model: The indoor temperature judgment model is fixed, meteorological parameters are sampled from noise data as input, and the cooling and heating load prediction model is used to output cooling and heating load prediction values; Compare the predicted cooling and heating loads with the actual cooling and heating loads and adjust the model parameters.

[0009] An electronic device comprises a processor and a memory which is in communication with the processor and is used to store instructions executable by the processor, wherein the processor is used to execute the above-mentioned method for predicting building cooling and heating loads based on a generative adversarial network.

[0010] A server includes at least one processor and a memory communicatively connected to the processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the processor so that the at least one processor executes the above-mentioned building cooling and heating load prediction method based on a generative adversarial network.

[0011] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned building cooling and heating load prediction method based on a generative adversarial network.

[0012] Compared with the prior art, the building cooling and heating load prediction method based on the generative adversarial network created by the present invention has the following advantages: 1. The method for predicting building cooling and heating loads based on a generative adversarial network created by the present invention improves the prediction accuracy of cooling and heating loads. It does not need to obtain cooling and heating load data as input parameters through energy consumption simulation software and typical meteorological data. It is trained based on measured historical data, continuously adjusts the generator model and judges its model, thereby improving the prediction accuracy.

[0013] 2. The method for predicting building cooling and heating loads based on generative adversarial networks created by the present invention improves the convenience of cooling and heating load prediction. Based on randomly generated data, it reduces the requirements on the quantity and quality of input data, expands the cooling and heating load data set of the air-conditioning system, and reduces the complexity of prediction.

[0014] 3. The building cooling and heating load prediction method based on the generative adversarial network created by the present invention takes into account the cumulative nature of system operation, and continuously updates the input data based on the historical data generated during the operation of the building air-conditioning system, thereby improving the diversity of load and temperature data, allowing the prediction model to continuously evolve and update, and continuously improving the prediction accuracy during operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings constituting part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments and descriptions of the present invention are used to explain the present invention and do not constitute an improper limitation on the present invention. In the drawings: Figure 1 A flowchart of a method for predicting building cooling and heating loads based on a generative adversarial network according to an embodiment of the present invention is shown; Figure 2A schematic diagram showing the conceptual architecture of a method for predicting building cooling and heating loads based on a generative adversarial network according to an embodiment of the present invention is shown; Figure 3 The schematic diagram of the architecture of the building cooling and heating load prediction method based on the generative adversarial network described in the embodiment of the invention is shown; Figure 4 A schematic diagram of the implementation and operation of the building cooling and heating load prediction method based on a generative adversarial network described in an embodiment of the invention is shown. DETAILED DESCRIPTION

[0016] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0017] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0018] The present invention proposes a building cold and hot load prediction method based on a generative adversarial network, the method comprising: constructing a building cold and hot negative prediction model; constructing a building indoor temperature judgment model including cold and hot load data as input; randomly generating building cold and hot load prediction data through the cold and hot negative prediction model, and mixing with cold and hot load real data to input into the indoor temperature judgment model; comparing the indoor predicted temperature obtained by the indoor temperature judgment model with the indoor real temperature to judge the authenticity of the input cold and hot load value; through continuous alternating cycle training, the accuracy of the building cold and hot load prediction model and the building indoor temperature judgment model is repeatedly improved; the air-conditioning system operation data is continuously used as a supplement to the database, and the cycle training is used to realize model evolution, so as to further improve the prediction accuracy.

[0019] like Figure 1-Figure 4 As shown, the building cooling and heating load prediction method S100 based on the generative adversarial network includes the following steps: Step S110, constructing a cooling and heating load prediction model and an indoor temperature judgment model; the indoor temperature judgment model can use a mixture of cooling and heating load real data and cooling and heating load prediction data output by the cooling and heating load prediction model as input to obtain indoor predicted temperature.

[0020] Before building the model, the historical data of the cooling and heating loads of the building (i.e., the air-conditioning system), the historical data of the indoor temperature, and the historical data of the meteorological parameters are obtained and preprocessed.

[0021] Among them, meteorological parameters include outdoor temperature and humidity, solar radiation intensity, and wind speed.

[0022] The cooling and heating load values ​​of the building are calculated by the air conditioning side parameters, which include the supply water temperature tg, the return water temperature th, and the supply and return water flow rate m. Let c be the specific heat capacity constant of water, and the calculation formula of the building cooling and heating load Qi at time i is as follows: Qi = cm(t g -t h ) Preprocessing of historical data: outliers are removed through feature engineering cleaning and denoising, and missing values ​​are supplemented. The normalization formula is used to unify the dimension of the data, and a standardized data set is constructed to obtain a historical database of cooling and heating loads, a historical database of indoor temperature, and a historical database of meteorological parameters.

[0023] For example, the operating data of the air-conditioning system, including supply and return water temperature, flow rate and other parameters, can be automatically recorded through the building management system (BMS); indoor temperature data of each area can be collected using indoor temperature sensors; and outdoor meteorological parameters corresponding to the building location can be obtained from public meteorological data services.

[0024] Use data cleaning techniques to remove obvious false readings and outliers. For example, use the Z-Score method to identify and remove outliers. Fill missing data with the average of the previous and next time points. Normalize all collected data to ensure that the data is at the same level for easy model processing. Use the minimum-maximum normalization method to scale the data between 0 and 1. The normalization formula is as follows: x' = \frac{x - \min(x)}{\max(x) - \min(x)} Where (x) is the original data and (x') is the normalized data.

[0025] In this embodiment, the method for constructing the cold and hot load prediction model is: A one-dimensional convolutional neural network (1D-CNN) is used to calculate the cooling and heating load values ​​based on the air conditioning side parameters to build a cooling and heating load prediction model; The historical data of meteorological parameters are used as input and the predicted values ​​of cold and hot loads are used as output to train the cold and hot load prediction model.

[0026] Specifically, a one-dimensional convolutional neural network (1D-CNN) is selected as the basic architecture of the cold and hot load prediction model. Considering its advantages in processing time series data, it can effectively capture the local characteristics of time series data. The specific structure of the model is set, including the number of layers, the number of neurons in each layer, the type of activation function, etc. For the cold and hot load prediction model, two convolutional layers are selected, each followed by a maximum pooling layer, and two fully connected layers. The time series data is windowed, and each window contains data from N consecutive time points to meet the input requirements of the 1D-CNN model.

[0027] Model training, using deep learning frameworks such as Keras to build a 1D-CNN model. Define the loss function as mean square error (MSE), and select Adam as the optimizer. Divide the preprocessed dataset into a training set and a validation set with a ratio of 8:2. Use the training set data to train the model, evaluate the model performance on the validation set, and adjust the model parameters until the prediction error on the validation set is minimized.

[0028] Model optimization and parameter adjustment: adjust the model parameters according to the performance on the validation set. This includes adjusting the number of convolutional layers, the number of neurons in each layer, and selecting different activation functions (such as ReLU, tanh, etc.). Implement early stopping to avoid overfitting. Stop training when the performance of the model on the validation set does not improve significantly after multiple consecutive training cycles (epochs). Use the trained model to predict the test set (or real-time data), and the output cold and hot load prediction data is used as the input of the subsequent indoor temperature judgment model.

[0029] In this embodiment, the method for constructing the indoor temperature judgment model is: Use one-dimensional convolutional neural network (1D-CNN) to build an indoor temperature judgment model; The indoor temperature judgment model is trained by taking the historical data of cooling and heating loads and historical data of meteorological parameters as input and the predicted indoor temperature as output.

[0030] Specifically, 1D-CNN is selected as the architecture of the indoor judgment model, but the Sigmoid activation function is used in the output layer to determine the probability that the input cold and hot load data is real data. The input data of the indoor temperature judgment model includes historical data from cold and hot loads and historical data from meteorological parameters.

[0031] For model training, the loss function is defined as binary cross entropy, and the optimizer is still Adam. Training is also performed on the training set, and the performance is evaluated on the validation set. Parameters are adjusted according to the performance of the model on the validation set, including adjusting the network structure, activation function, etc., and early stopping is used to prevent overfitting. The test set is used to evaluate the model performance, and the main indicator of concern is accuracy, that is, the proportion of correct judgments on the authenticity of the input load data.

[0032] Step S120, based on the generative adversarial network, alternately train the cold and hot load prediction model and the indoor temperature judgment model to obtain the target cold and hot load prediction model; the indoor temperature judgment model judges the authenticity of the input cold and hot load data by distinguishing the authenticity of the indoor predicted temperature and the indoor actual temperature, thereby optimizing its model parameters.

[0033] The alternating training method of the cold and hot load prediction model and the indoor temperature judgment model is: Training indoor temperature judgment model: The cooling and heating load prediction model is fixed, the real cooling and heating load data and the cooling and heating load prediction data output by the cooling and heating load prediction model are mixed as input, and the indoor temperature prediction model is used to obtain the indoor temperature. Compare the predicted indoor temperature with the actual indoor temperature, and judge the authenticity of the predicted indoor temperature and the actual indoor temperature, as well as the authenticity of the actual data of the cooling and heating loads and the predicted data of the cooling and heating loads; That is, the indoor temperature judgment model inputs a real load or a false load, and compares the output indoor predicted temperature with the indoor real temperature. The indoor temperature judgment model determines which of the two temperatures is real, and then determines whether the input load is real or false.

[0034] Calculate the loss function and update the model parameters based on the judgment results.

[0035] In this embodiment, two types of load data may be labeled (the cold and hot load prediction data is 0, and the cold and hot load actual data is 1).

[0036] Training the cooling and heating load prediction model: The indoor temperature judgment model is fixed, meteorological parameters are sampled from noise data as input, and the cold and hot load prediction model is used to output the cold and hot load prediction values; Compare the predicted cooling and heating loads with the actual cooling and heating loads and adjust the model parameters.

[0037] The training steps are repeated alternately to continuously optimize the cold and hot load prediction model and the indoor temperature judgment model, and to continuously improve the accuracy of the cold and hot load prediction model.

[0038] Loss calculation and optimization: Compare the predicted indoor temperature with the actual indoor temperature, use the loss function (such as mean square error MSE) to calculate the loss, and update the parameters of the indoor temperature judgment model. Sample meteorological parameters from the noise data and input them into the cold and hot load prediction model to generate cold and hot load prediction values. Fix the indoor temperature judgment model, and adjust the parameters of the indoor temperature judgment model according to the judgment results of the indoor temperature judgment model on the true and false loads to improve the "authenticity" of the false loads it generates. Alternate the training process of the indoor temperature judgment model and the cold and hot load prediction model, set evaluation indicators (such as the difference between the generated data and the real data), monitor the training process, ensure that the model is optimized in the right direction, and gradually improve the performance of the two models.

[0039] Moreover, this method can collect new cold and hot load data, meteorological parameter data, and indoor temperature data from the air conditioning system in real time. Preprocess the newly collected data, including cleaning and normalization. Add new data to the training set and continuously expand the database. Repeat the model alternating training steps using the updated data set to continuously train and optimize the cold and hot load prediction model and the indoor temperature judgment model. Regularly evaluate the prediction accuracy of the model to ensure the reliability of the prediction results. According to the performance evaluation results, refine and adjust the model parameters to adapt to data changes and improve the prediction accuracy.

[0040] That is, the air-conditioning system operation data is used as a supplement to the historical database, and the newly generated cold and hot load data and meteorological parameter data are used as the input of the cold and hot load prediction model. New cold and hot load prediction values ​​are generated as the input of the indoor temperature judgment model, and compared with the newly generated indoor temperature data. The cold and hot load prediction model is continuously corrected through the newly generated data to achieve continuous evolution of the cold and hot load prediction model, further improve the prediction accuracy, and use it to guide the control of the building air-conditioning system.

[0041] Step S130, predicting the cooling and heating loads of the building according to the target cooling and heating load prediction model to obtain a predicted cooling and heating load value of the building.

[0042] Refer to the attached Figure 4 As shown, this embodiment takes an office building as an example to explain in detail the building cooling and heating load prediction method based on the generative adversarial network. The specific steps are as follows: Model deployment and real-time prediction and control: Use sensors to measure the supply and return water temperature and flow of air conditioning units and record data regularly. Automatically collect data from each sensor through the building management system and upload it to the central database in real time. Perform preliminary cleaning of the collected data, including removing obvious erroneous data and outliers. Establish a historical database for hot and cold loads, store the cleaned data in a dedicated database, and mark each record with a timestamp for subsequent processing. Install temperature sensors at key locations in the building to monitor the indoor temperature of each area in real time. Also automatically collect indoor temperature data through the building management system and upload it to the central database.

[0043] The historical data of cold and hot loads and indoor temperature are used to train two cold and hot load prediction models and an indoor temperature judgment model, and the model is constructed through a deep learning framework. The model is preliminarily trained and the network parameters are adjusted until the prediction error of the model is reduced to an acceptable range, thus obtaining a cold and hot load prediction model and an indoor temperature judgment model.

[0044] Deploy the two trained models to the central control platform of the building air conditioning system, and train the two models alternately according to S120 to ensure that the models can accurately predict the cooling and heating loads. Automatically collect data during the daily operation of the building air conditioning system, and regularly (such as once a week) retrain the model with the newly collected data to adapt to changes in the environment and system operation. Analyze the accuracy of the model prediction, and adjust the model structure or training method as needed.

[0045] Collect the current meteorological parameters and indoor environmental parameters as model inputs. Use the trained model to calculate the predicted cooling and heating loads and predicted indoor temperature at the next moment. Compare the predicted results with the actual operating data to evaluate the accuracy of the prediction. Feed the newly generated data during the system operation back to the model for continuous online learning and optimization to ensure that the model's prediction ability gradually improves over time and adapts to environmental changes and changes in building usage patterns.

[0046] Upload the predicted cooling and heating load data to the central control platform of the building air conditioning system. Based on the predicted data, the central control platform calculates the optimal operating parameters of the air conditioning unit, such as supply and return water temperature, flow rate, etc. Send control commands to the air conditioning unit to adjust its operating status to meet the predicted load demand. Monitor the actual operating effect of the air conditioning system, including energy consumption, indoor temperature stability, etc. Feedback the operating effect to the model to further optimize the prediction accuracy. Adjust the control strategy of the central control platform based on the feedback results to achieve more refined energy management and control to ensure indoor comfort.

[0047] Those of ordinary skill in the art will appreciate that the units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0048] In the several embodiments provided in the present application, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the division of the units described above is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The above-mentioned units may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0049] 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

[0050] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A building cooling and heating load prediction method based on generative adversarial network, characterized in that: The steps include: Constructing a cold and hot load prediction model and an indoor temperature judgment model; the indoor temperature judgment model can use a mixture of cold and hot load real data and cold and hot load prediction data output by the cold and hot load prediction model as input to obtain an indoor predicted temperature; Based on the generative adversarial network, the cooling and heating load prediction model and the indoor temperature judgment model are alternately trained to obtain a target cooling and heating load prediction model; the indoor temperature judgment model judges the authenticity of the input cooling and heating load data by judging the authenticity of the indoor predicted temperature output by it and the indoor real temperature, thereby optimizing its model parameters; The cooling and heating loads of the building are predicted according to the target cooling and heating load prediction model to obtain the cooling and heating load prediction value of the building.

2. The method for predicting building cooling and heating loads based on a generative adversarial network according to claim 1 is characterized in that: The method for constructing the cooling and heating load prediction model is: Using a one-dimensional convolutional neural network, the cooling and heating load values ​​are calculated based on the air-conditioning side parameters to build an initial cooling and heating load prediction model; the air-conditioning side parameters include supply water temperature, return water temperature, and supply water return water flow rate; The cooling and heating load prediction model is obtained by training using historical data of meteorological parameters as input and predicted values ​​of cooling and heating loads as output; the meteorological parameters include outdoor temperature and humidity, solar radiation intensity, and wind speed.

3. The building cooling and heating load prediction method based on generative adversarial network according to claim 1 is characterized in that: The method for constructing the indoor temperature judgment model is: Use one-dimensional convolutional neural network to build indoor temperature judgment model; The indoor temperature judgment model is obtained by training with historical data of cooling and heating loads and historical data of meteorological parameters as input and predicted indoor temperature as output.

4. The method for predicting building cooling and heating loads based on a generative adversarial network according to claim 1 is characterized in that: The alternating training method of the cooling and heating load prediction model and the indoor temperature judgment model is: Training the indoor temperature judgment model: The cooling and heating load prediction model is fixed, the real cooling and heating load data and the cooling and heating load prediction data output by the cooling and heating load prediction model are mixed as input, and the indoor temperature prediction model is used to obtain the indoor temperature; Compare the predicted indoor temperature with the actual indoor temperature, and judge the authenticity of the predicted indoor temperature and the actual indoor temperature, as well as the authenticity of the actual data of the cooling and heating loads and the predicted data of the cooling and heating loads; Calculate the loss function and update the model parameters based on the judgment results; Training the cooling and heating load prediction model: The indoor temperature judgment model is fixed, meteorological parameters are sampled from noise data as input, and the cooling and heating load prediction model is used to output cooling and heating load prediction values; Compare the predicted cooling and heating loads with the actual cooling and heating loads and adjust the model parameters.

5. An electronic device, comprising a processor and a memory connected to the processor for storing instructions executable by the processor, characterized in that: The processor is used to execute the building cooling and heating load prediction method based on the generative adversarial network as described in any one of claims 1 to 4.

6. A server, characterized in that: It includes at least one processor and a memory communicatively connected to the processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the processor so that the at least one processor executes the building cooling and heating load prediction method based on a generative adversarial network as described in any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the building cooling and heating load prediction method based on a generative adversarial network described in any one of claims 1 to 4 is implemented.

Images