A simulation method and system for testing the energy efficiency of a water heater

Abstract

This invention discloses a simulation method and system for testing the energy efficiency of water heaters, belonging to the field of simulation technology. The method includes: constructing a parameterized unit library for the water heater using its design drawings and component characteristics; synchronously driving the simulation process of the parameterized unit library and a physical water heater prototype under preset basic test conditions; iteratively comparing the measured data stream of the physical prototype with the simulation output data stream to form a high-fidelity digital twin; collecting historical user water usage data to construct a historical user water usage behavior dataset; training the historical user water usage behavior dataset using a generative adversarial network; and using the trained generator network to synthesize a virtual user water usage behavior time series representing extreme water usage patterns. This invention drives the high-fidelity digital twin to run under extreme conditions, enabling the calculation of its overall energy efficiency under harsh conditions and the generation of an evaluation report.

Description

Technical Field

[0001] This invention relates to the field of simulation technology, and in particular to a simulation method and system for testing the energy efficiency of water heaters. Background Technology

[0002] With the deepening of energy strategies and market attention to the energy efficiency of home appliances, the energy efficiency of water heaters has become a core competitive indicator. Energy efficiency assessment mainly relies on bench testing of physical prototypes in the laboratory according to national standards. This method is authoritative but costly and time-consuming. Virtual testing technology based on simulation can be applied. By establishing a theoretical model of the water heater and inputting standard operating conditions, its working process can be simulated to predict energy efficiency, providing an effective auxiliary tool for research and development and reducing the cost of trial and error in the early stage.

[0003] Existing simulation methods have limitations in model accuracy and operating condition coverage. Regarding model accuracy, simulations are typically based on general theoretical models and empirical parameters, making it difficult to accurately reflect individual differences caused by manufacturing tolerances and component performance variations. This results in inherent deviations between simulation results and measured data from specific physical prototypes, affecting the reliability of simulations as a basis for high-confidence design. In terms of operating condition coverage, simulation inputs are mostly limited standard or typical operating conditions, making it difficult to realistically simulate the randomness, diversity, and extreme modes of user water usage. This means that simulation evaluations cannot fully reveal the performance boundaries and potential risks of products under complex, high-pressure conditions. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a simulation method for testing the energy efficiency of water heaters, which solves the problem that existing simulation methods cannot perform high-confidence assessments of the true performance boundaries and reliability of products due to insufficient model individualization fidelity and incomplete coverage of extreme test conditions.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a simulation method for testing the energy efficiency of a water heater, which includes constructing a parameterized unit library for the water heater using the design drawings and component characteristics of the water heater; Under the preset basic test conditions, the simulation process that drives the water heater parameterized unit library to run synchronously is compared with the physical water heater prototype. The measured data stream of the physical prototype is iteratively compared with the simulation output data stream to form a high-fidelity digital twin. Collect historical user water usage data to construct a historical user water usage behavior dataset; Generative adversarial networks are used to train a dataset of historical user water use behavior, and the trained generator network is used to synthesize time series of virtual user water use behavior with extreme water use patterns. Based on the time series of virtual user water usage behavior, a high-fidelity digital twin is driven to run, calculate the overall energy efficiency value of the water heater, and generate an energy efficiency and evaluation report of the water heater under extreme operating conditions.

[0007] As a preferred embodiment of the simulation method for water heater energy efficiency testing according to the present invention, the construction of the water heater parameterized unit library includes, By utilizing the design drawings and component characteristics of the water heater, the functional components that make up the water heater are identified, and a mathematical relationship description of the water heater components is established. In the mathematical relationship description of water heater components, quantifiable physical quantities that affect energy efficiency are set as undetermined parameters in the mathematical relationship description of water heater components, and standardized sets of input and output variables are formed, which are then integrated into a water heater parameterized unit library.

[0008] As a preferred embodiment of the simulation method for testing the energy efficiency of a water heater according to the present invention, the simulation process includes, A set of basic test conditions is preset. The parameterized unit library of the water heater is used to parse and match the instruction sequence of the basic test conditions and input variables. The calculation step size is set to balance the simulation accuracy and speed, thus forming the simulation process. Under the preset basic test conditions, the same basic test condition commands are converted into physical control signals executed by the physical water heater prototype through the measurement and control equipment, thereby driving the physical water heater prototype to operate.

[0009] As a preferred embodiment of the simulation method for water heater energy efficiency testing according to the present invention, the high-fidelity digital twin includes, Collect the measured data stream of the physical prototype water heater during operation; obtain the simulation output data stream generated by the simulation process by reading the values ​​of virtual monitoring points during the simulation process. The measured data stream of the physical prototype is iteratively compared with the simulation output data stream. Based on the iterative comparison results of the measured data stream of the physical prototype and the simulation output data stream, the undetermined parameters in the parameterized unit library of the water heater are driven to evolve autonomously. When the dynamic response of the simulation process converges with the actual operating response of the physical water heater prototype within the preset tolerance range, the state of the water heater parameterized unit library is recorded to form a high-fidelity digital twin.

[0010] As a preferred embodiment of the simulation method for water heater energy efficiency testing described in this invention, the historical user water usage behavior dataset includes, Collect historical user water usage data from smart water heaters and perform cleaning and formatting processes; The historical user water usage data, after being cleaned and formatted, is integrated into a historical user water usage behavior dataset.

[0011] As a preferred embodiment of the simulation method for water heater energy efficiency testing described in this invention, the training of the historical user water usage behavior dataset includes: Time series samples are extracted from historical user water usage behavior datasets, and the time series samples are input into a generative adversarial network for adversarial training to obtain a generator network and a discriminator network. The generator network receives random noise vectors to obtain simulated time series, and the discriminator network receives real time series samples and performs authenticity judgment against the simulated time series generated by the generator network. During adversarial training, the generator network and the discriminator network engage in iterative game and parameter updates using time-series samples as a reference.

[0012] As a preferred embodiment of the simulation method for water heater energy efficiency testing according to the present invention, the virtual user water usage behavior time series includes, The generator network adjusts its internal parameters based on the feedback signal of the self-discriminator network, so that the simulated time series obtained by the generator network is close to the real time series samples in the historical user water use behavior dataset in terms of statistical characteristics, time patterns and numerical distribution, thereby enabling the generator network to generate water use behavior sequences from random noise. By inputting a guiding noise vector into the trained generator network The guiding noise vector points to the region of high intensity and high volatility in the distribution learned by the generator network, driving the generator network to output a time series of virtual user water usage behavior.

[0013] As a preferred embodiment of the simulation method for testing the energy efficiency of a water heater according to the present invention, the calculation of the overall energy efficiency value of the water heater includes, The time series of virtual user water usage behavior is loaded into a high-fidelity digital twin, and the operation of the high-fidelity digital twin generates high-fidelity digital twin operation process data. Energy consumption and effective heat data are extracted from the high-fidelity digital twin's operation data, and the overall energy efficiency value of the water heater is calculated using the energy consumption and effective heat data.

[0014] As a preferred embodiment of the simulation method for water heater energy efficiency testing according to the present invention, the energy efficiency and evaluation report includes: The time series of performance indicators are extracted from the high-fidelity digital twin's operation data, and the time series of performance indicators are comprehensively analyzed in combination with the overall energy efficiency value of the water heater. Based on the comprehensive analysis results, an energy efficiency and evaluation report of the water heater under extreme operating conditions is generated.

[0015] Secondly, the present invention provides a simulation system for testing the energy efficiency of a water heater, comprising a construction module for constructing a parameterized unit library for the water heater using the design drawings and component characteristics of the water heater; The comparison module, under preset basic test conditions, synchronously drives the simulation process of the water heater parameterized unit library and the physical water heater prototype, and iteratively compares the measured data stream of the physical prototype with the simulation output data stream to form a high-fidelity digital twin. The data acquisition module collects historical user water usage data and constructs a historical user water usage behavior dataset. The training module uses a generative adversarial network to train on a historical user water use behavior dataset, and uses the trained generator network to synthesize virtual user water use behavior time series with extreme water use patterns. The assessment report module, based on the time series of virtual user water usage behavior, drives the operation of a high-fidelity digital twin to calculate the overall energy efficiency value of the water heater and generate an energy efficiency and assessment report of the water heater under extreme operating conditions.

[0016] The beneficial effects of this invention are as follows: By combining digital twin generation with Generative Adversarial Networks (GANs), efficient and high-precision virtual evaluation of water heater performance under extreme operating conditions is achieved. By constructing a parameterized component library and utilizing synchronous testing and iterative comparison, the simulation model parameters are driven to evolve autonomously, generating a high-fidelity digital twin that is highly consistent with a specific physical prototype. This solves the problem that traditional simulation models, due to their generalization, cannot accurately match the characteristics of individual physical objects, ensuring the accuracy of the virtual test object. Furthermore, by using GANs to train on historical water usage data, extreme water usage behavior sequences containing characteristics such as high intensity and high volatility are synthesized, overcoming the limitations of relying on limited standard operating conditions. This achieves automatic coverage of complex and high-pressure usage scenarios, driving the high-fidelity digital twin to operate under extreme conditions, allowing for the calculation of its overall energy efficiency under harsh conditions and the generation of an evaluation report. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a simulation method for testing the energy efficiency of water heaters.

[0019] Figure 2 This is a schematic diagram of a simulation system for testing the energy efficiency of a water heater. Detailed Implementation

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "example" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the invention. The appearance of an embodiment in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments.

[0023] Reference Figures 1-2 As one embodiment of the present invention, this embodiment provides a simulation method for testing the energy efficiency of a water heater, comprising the following steps: S1. Construct a parametric unit library for water heaters using the design drawings and component characteristics of the water heater.

[0024] S1.1 Using the design drawings and component characteristics of the water heater, identify the functional components that make up the water heater and establish a mathematical relationship description of the water heater components.

[0025] Furthermore, based on the design drawings of the water heater, independent functional components such as the burner, heat exchanger, water pump, water tank, temperature sensor, and proportional valve are identified. For each identified functional component, a mathematical relationship description of its behavior is established based on its physical working principle. For example, a description of the combustion heat release rate based on chemical kinetics and mass conservation is established for the burner; a description of heat transfer based on convection and conduction is established for the heat exchanger; a head-flow characteristic curve description is established for the water pump; and a proportional-integral-derivative regulation law description is established for the controller.

[0026] S1.2 In the mathematical relationship description of water heater components, quantifiable physical quantities that affect energy efficiency are set as undetermined parameters in the mathematical relationship description of water heater components, and standardized sets of input and output variables are formed, which are then integrated into a water heater parameterized unit library.

[0027] Furthermore, in each established mathematical relationship description of a water heater component, physical quantities that determine the component's efficiency or performance and fluctuate due to manufacturing processes or operating conditions, such as the convective heat transfer coefficient of the heat exchanger, the friction coefficient of the pipeline, the gain and zero-point drift of the sensor, and the proportional gain of the controller, are explicitly designated as undetermined parameters in the mathematical relationship description of the water heater component. For each mathematical relationship description of a water heater component containing undetermined parameters, a standardized set of input variables necessary for its interaction with the external environment or other components is defined, such as the hot water flow rate and inlet temperature entering the heat exchanger, as well as a standardized set of output variables, such as the hot water temperature and pressure drop exiting the heat exchanger. All mathematical relationship descriptions of water heater components that have completed the above definitions, have standardized input and output interfaces, and whose key parameters are undetermined are summarized and managed to form a water heater parameterized unit library.

[0028] S2. Under the preset basic test conditions, the simulation process of the water heater parameterized unit library is synchronously driven to run with the physical water heater prototype. The measured data stream of the physical prototype is iteratively compared with the simulation output data stream to form a high-fidelity digital twin.

[0029] S2.1. Preset a set of basic test conditions, use the water heater parameterized unit library to parse and match the input variables of the instruction sequence of the basic test conditions, set the calculation step size to balance the simulation accuracy and speed, and form a simulation process.

[0030] Furthermore, working point combinations are defined to constitute basic test conditions. The construction of basic test conditions is achieved by selecting several different inlet water temperature values ​​and several different outlet water flow rates. For example, three inlet water temperatures covering low, normal, and slightly high temperatures are selected, along with three outlet water flow rates covering low, medium, and slightly high flow rates. Each inlet water temperature is paired with each outlet water flow rate to generate a condition matrix covering multiple working areas. The water heater parameterized unit library is used to parse and match the input variables of the instruction sequences for the basic test conditions. Each pair of temperature and flow parameters is converted into a step or ramp instruction sequence containing time information. The temperature value in the instruction sequence is assigned to the temperature input variable in the standardized input variable set of the corresponding inlet unit in the water heater parameterized unit library, and the flow rate value is assigned to the flow rate input variable in the standardized input variable set of the corresponding water pump or valve unit. Setting a calculation step size to balance simulation accuracy and speed involves configuring a fixed time step size for the solver of the water heater parameterized unit library. This requires ensuring the numerical stability of the solution process and capturing the main frequency components of the water heater's dynamic response, forming a simulation process driven by the parsed instruction sequence.

[0031] Specifically, different inlet water temperatures and outlet water flow rates are fully or partially paired to generate basic test conditions that actively stimulate the water heater's operating state under different load rates and heat transfer temperature differences. For example, the combination of low inlet water temperature and high flow rate tests the heater's maximum output capacity and transient response, while the combination of high inlet water temperature and low flow rate tests the stability and minimum adjustment capability of the control system. This ensures that the simulation process can traverse various dynamic modes of the water heater's parameterized unit library within its main operating range, providing structured input stimuli for obtaining comprehensive and sufficient comparison data in subsequent steps, which is a prerequisite for effective model calibration.

[0032] S2.2 Under the preset basic test conditions, the same basic test condition commands are converted into physical control signals executed by the physical water heater prototype through the measurement and control equipment, thereby driving the physical water heater prototype to run.

[0033] Furthermore, under the preset basic test conditions, an identical sequence of instructions constituting the basic test conditions is output to the measurement and control equipment. Upon receiving the instruction sequence, the measurement and control equipment converts the digital temperature and flow setpoints into continuous analog voltage or current signals via a digital-to-analog converter. These signals are used to drive the adjustable valves or variable frequency pumps in the prototype water heater. Based on the timing logic in the instruction sequence, the measurement and control equipment controls the power relays or contactors to manage the start and stop of the heating elements in the prototype water heater. The basic test condition instructions are precisely converted into physical control signals executable by the prototype water heater. When driving the prototype water heater, it is necessary to ensure that the timing, amplitude, and duration of the physical control signals output by the measurement and control equipment are completely synchronized with the instruction sequence received by the simulation process, enabling the prototype water heater to operate repeatedly under the preset basic test conditions.

[0034] Specifically, the same digital operating condition commands are fed seamlessly into both the virtual simulation process and the physical water heater prototype. For example, when the command sequence requires the flow rate to jump from a low value to a high value at a specific moment, the simulation process achieves this by updating internal variables, while the physical prototype achieves it by driving the valve opening change through measurement and control equipment. Both respond to the same command at the same moment, eliminating contrast noise introduced by asynchronous or different sources of excitation signals. This is attributed to the degree of consistency between the dynamic characteristics of the water heater parameterized unit library and the actual physical characteristics of the physical water heater prototype.

[0035] S2.3. Collect the measured data stream of the physical prototype generated by the operation of the physical water heater prototype; obtain the simulation output data stream generated by the simulation process by reading the values ​​of the virtual monitoring points during the simulation process. Furthermore, the actual measured data stream generated by the operation of the physical water heater prototype is acquired by continuously recording measured values ​​at a constant high sampling frequency using physical sensors, including temperature sensors, pressure sensors, flow meters, and gas analyzers, connected to the water inlet, outlet, combustion chamber, heating element, and exhaust port of the physical water heater prototype. This forms a multi-channel physical quantity measurement sequence that is strictly synchronized with time; this sequence is the actual measured data stream of the physical prototype. The simulation output data stream is obtained by reading the values ​​of virtual monitoring points during the simulation process. Virtual computing nodes corresponding to the installation locations of the aforementioned physical sensors in the water heater parameterization unit library read and record the calculated values ​​of each physical quantity in real time at the same sampling frequency as the physical data acquisition, forming a multi-channel physical quantity calculation sequence that is completely aligned with the actual measured data stream of the physical prototype in the time dimension; this sequence is the simulation output data stream.

[0036] Specifically, based on the mirror monitoring principle, a one-to-one mapping relationship is established between the key performance monitoring points of the physical prototype and the corresponding computational nodes in the simulation process, and the same sampling strategy is enforced for data recording. For example, a temperature sensor is placed at the outlet of the physical prototype, that is, a virtual temperature monitoring point is set at the end unit of the water circuit in the simulation process. This ensures that the information carried by the measured data stream of the physical prototype and the simulation output data stream is completely comparable in physical meaning, time coordinate, and data dimension, thereby transforming the complex problem of overall performance comparison into a problem of direct comparison of time series data on multiple channels.

[0037] S2.4. Iteratively compare the measured data stream of the physical prototype with the simulation output data stream.

[0038] Furthermore, the measured data stream from the physical prototype and the simulation output data stream are strictly aligned according to timestamps. Then, for each corresponding physical quantity data sequence, such as the effluent temperature sequence, a statistical measure of the difference between the measured and simulated values ​​over the entire time range is calculated, such as root mean square error or mean absolute error. This process can calculate the error individually for each sub-condition in the basic test conditions, and then combine the errors of all sub-conditions to obtain an overall difference evaluation index. After one comparison is completed, a new round of driving, data acquisition, and comparison can be triggered as needed, constituting an iteration.

[0039] Specifically, the comparison object is expanded from static operating point results to dynamic, complete response curves covering multiple operating conditions. By calculating statistical measures such as root mean square error, the overall deviation of the simulation process from the behavior of the physical prototype in both transient and steady-state phases can be comprehensively evaluated, rather than focusing solely on the error at a specific moment. The comparison strategy based on the overall similarity of time series can more sensitively and comprehensively capture the model's shortcomings in dynamic characteristics, such as response delay, overshoot, or oscillation trends.

[0040] S2.5 Based on the iterative comparison results of the measured data stream of the physical prototype and the simulation output data stream, drive the undetermined parameters in the parameterized unit library of the water heater to evolve autonomously; Furthermore, the evolution process uses the overall difference evaluation index as the optimization objective function. A parameter optimization algorithm is employed to automatically adjust the values ​​of individual or grouped undetermined parameters in the water heater parameterized unit library, re-executing the entire process from driving the water heater parameterized unit library to calculating the difference evaluation index. The operating logic of the optimization algorithm is to find a specific set of undetermined parameter assignments that minimizes the overall difference evaluation index between the simulation output data stream generated by running the simulation process under this set of parameters and the actual measured data stream of the physical prototype.

[0041] Specifically, the goal of reducing the overall difference evaluation index is transformed into an automated optimization problem using undetermined parameters in the water heater parameterized unit library as decision variables. By embedding a parameter optimization algorithm, this method can automatically explore the space of undetermined parameters and intelligently adjust parameter values ​​based on the feedback of the comparison results of each iteration, driving the dynamic response of the simulation process to approximate the actual operating response of the physical prototype. For example, if the error in the outlet water temperature response delay is large, the optimization algorithm may automatically increase the value of the undetermined parameter that affects the heat transfer rate.

[0042] S2.6 When the dynamic response of the simulation process and the actual operating response of the physical water heater prototype converge within the preset tolerance range, record the state of the water heater parameterized unit library to form a high-fidelity digital twin.

[0043] Furthermore, a preset tolerance range is set as the convergence criterion. After each execution of parameter evolution and recalculation of the overall difference evaluation index, it is determined whether the index has fallen within the preset tolerance range. An additional check can be made to determine whether the decrease in the index over multiple iterations is negligible to confirm the stability of convergence. The iterative optimization process is terminated only when the difference between the dynamic response of the simulation process and the actual operating response of the physical water heater prototype is determined to be convergent within the preset tolerance range. The values ​​of the undetermined parameters, the connection topology between units, and the operating logic of all units in the current water heater parameterized unit library are recorded. The recorded complete state constitutes a high-fidelity digital twin corresponding to the current physical water heater prototype.

[0044] Specifically, the preset tolerance range provides a clear threshold for accuracy requirements, giving the calibration process a clear termination target and avoiding overfitting or infinite loops. Recording the state of the water heater parameterized unit library that meets the conditions essentially captures the model parameter set and structure that, after data calibration, best reproduces the dynamic behavior of the specific entity prototype. The resulting high-fidelity digital twin is no longer a general model, but an individualized virtual entity with identity significance, verified by measured data.

[0045] S3. Collect historical user water usage data and construct a historical user water usage behavior dataset.

[0046] S3.1 Collect historical user water usage data from smart water heaters and perform cleaning and formatting processes.

[0047] Furthermore, through network connections or data interfaces, historical user water usage data is retrieved in batches from smart water heaters deployed at the user's end or their associated data platforms. This historical user water usage data typically includes raw records of fields such as timestamps, hot water flow rates, set temperatures or actual outlet water temperatures, and possibly heating power or gas flow rates. The historical user water usage data undergoes cleaning and formatting processes, including identifying and removing outliers that clearly exceed physically reasonable ranges, such as negative flow rates or extremely high temperatures; handling data loss due to communication interruptions or device dormancy using interpolation or marker deletion methods; and unifying and aligning data from different smart water heaters or different periods in terms of timestamps, units of measurement, and data precision to form a temporally continuous, physically consistent, and formatted data sequence, thus completing the data cleaning and formatting process.

[0048] S3.2 Integrate the historical user actual water usage data after cleaning and formatting into a historical user water usage behavior dataset.

[0049] Furthermore, the historical user water usage data, after being cleaned and formatted, undergoes an integration process. This integration involves aggregating data sequences from multiple smart water heaters, users, and time periods, and organizing and managing them according to a unified data structure. During integration, it may be necessary to add metadata tags to each data record, such as the season, geographical location, or user group classification. Finally, the integrated, high-quality dataset with a unified structure is stored in a readily accessible and analyzable dataset format; this dataset constitutes the historical user water usage behavior dataset.

[0050] S4. Use a generative adversarial network to train on a historical user water use behavior dataset, and use the trained generator network to synthesize virtual user water use behavior time series with extreme water use patterns.

[0051] S4.1 Extract time series samples from the historical user water usage behavior dataset, input the time series samples into the generative adversarial network for adversarial training, and obtain the generator network and the discriminator network.

[0052] Furthermore, several consecutive time-series samples are extracted from the historical user water use behavior dataset according to certain rules. These time-series samples are sequence segments with fixed time lengths, containing multi-dimensional parameters such as flow rate and temperature, representing a complete or partial cycle of the historical user's actual water use behavior. These time-series samples extracted from the historical user water use behavior dataset are used as training data to prepare input for the Generative Adversarial Network (GAN). The GAN initialization includes a generator network and a discriminator network. The adversarial training process begins by inputting the time-series samples into the discriminator network in batches, while the generator network receives randomly generated noise vectors and generates simulated time-series outputs through internal computation. The generator and discriminator networks learn and update through adversarial training, with the ultimate goal of obtaining a well-trained generator network capable of generating realistic time-series data and a discriminator network with discriminative capabilities.

[0053] Specifically, by utilizing the game mechanism between the generator network and the discriminator network, the high-dimensional, nonlinear joint probability distribution behind real water usage data is learned indirectly and implicitly. Extracting time-series samples from historical user water usage behavior datasets provides real data anchors for adversarial training. Initializing the generator and discriminator networks and initiating adversarial training creates a dynamic learning environment. The generator network aims to produce simulated sequences sufficient to deceive the discriminator, forcing it to learn the complex correlations and statistical patterns among multiple dimensions of water usage events, such as timing, duration, and flow intensity, from real data.

[0054] S4.2 The generator network receives random noise vectors to obtain the simulated time series, and the discriminator network receives real time series samples and performs authenticity judgment on the simulated time series generated by the generator network.

[0055] Furthermore, the operation of the generator network and discriminator network in adversarial training is as follows: In each training iteration, the generator network receives one or more noise vectors randomly sampled from a preset distribution. These noise vectors undergo a nonlinear transformation by a multi-layered neural network within the generator network, mapping and reconstructing them into simulated time series with the same dimensional structure as the time series samples. Simultaneously, the discriminator network receives two types of input: real time series samples from historical user water usage behavior datasets, and the simulated time series generated by the generator network in this iteration. The discriminator network processes each input sequence, outputting a scalar value representing the probability that the sequence is classified as real. The discriminator network performs the task of distinguishing between real and simulated time series, striving to classify real time series samples as having a high probability and the simulated time series generated by the generator network as having a low probability, thus constituting a specific adversarial discrimination process.

[0056] Specifically, a clear dual-path contrastive learning mechanism is defined. The generator network generates structured sequences from unstructured random noise, its ability lying in its internal modeling of the real data distribution. The discriminator network simultaneously observes real samples and generated samples and performs binary classification, a discrimination process whose ability lies in distinguishing the boundaries of the data distribution. The closed-loop design of the generator-discriminator network forces the two networks to continuously compete and evolve. For example, the generator network may initially produce chaotic sequences, easily detected by the discriminator network; as training progresses, the generator network must learn to adjust its parameters to generate sequences that are statistically closer to real samples in order to improve the success rate of deception. This competitive mechanism is the core driving force that propels the generator network to ultimately learn a high-quality data distribution, ensuring that its output is not simple interpolation or noise, but a synthetic sequence that conforms to the statistical laws of real data.

[0057] S4.3 During adversarial training, the generator network and the discriminator network engage in iterative game and parameter updates using time-series samples as a reference.

[0058] Furthermore, during adversarial training, the generator network and the discriminator network use the input time-series samples as a common reference benchmark, engaging in multiple rounds of iterative game and parameter updates. Each iteration typically comprises two phases. In the first phase, the generator network's parameters are fixed, and the discriminator network's internal parameters are updated using the loss generated by its discrimination results against a batch of real samples and generated samples, through backpropagation, to improve the discriminator network's discrimination ability. In the second phase, the discriminator network's parameters are fixed again, and the generator network's internal parameters are updated using the loss generated when it fails to deceive the discriminator network with its simulated time-series samples, through backpropagation, to improve the generator network's ability to generate realistic sequences. The generator network and the discriminator network iterate and game alternately, with the goal of minimizing their respective adversarial losses. Their performance continuously competes and improves during this game until the training reaches a preset stopping condition, such as reaching the maximum number of iterations or the discriminator loss and generation loss reaching a dynamic equilibrium.

[0059] Specifically, the parameter optimization problem in machine learning is constructed as a minimax game with a clear adversarial objective. The generator network and the discriminator network compete against each other using time-series samples as the gold standard. This forces the generator network to explore the space of the data distribution, while the discriminator network continuously refines its characterization of the distribution boundaries. This iterative game mechanism, with alternating updates, ensures the stability and effectiveness of the learning process. The direction of parameter updates is always guided by the adversarial loss, making the evolution of the generator network's capabilities directly linked to the goal of generating sequences that can pass the test of the current strongest discriminator. The final dynamic equilibrium reached is when the sequences generated by the generator network are statistically indistinguishable from real time-series samples in the discriminator's eyes, signifying that the generator network has successfully mastered the core distribution characteristics of the historical user water use behavior dataset. S4.4 The generator network adjusts its internal parameters based on the feedback signal of the self-discriminator network, so that the simulated time series obtained by the generator network is close to the real time series samples in the historical user water use behavior dataset in terms of statistical characteristics, time patterns and numerical distribution, thereby enabling the generator network to generate water use behavior sequences from random noise.

[0060] Furthermore, in each training iteration, the generator network adjusts its internal trainable parameters based on the gradient signal fed back from the discriminator network. The gradient signal originates from the loss function of the discriminator network's judgment of the simulated time series generated by the generator network. Through backpropagation, the gradient of this loss function with respect to the generator network parameters is calculated and used to update the generator network's weights and biases. The direction of parameter adjustment aims to minimize the loss function, thereby maximizing the probability that the discriminator network misclassifies the simulated time series output by the generator network as real. By repeatedly executing this process, the adjustment of the generator network's internal parameters gradually makes the simulated time series obtained by the generator network increasingly closer to the real time series samples in the historical user water use behavior dataset in terms of multidimensional statistical characteristics, time dependency patterns, and numerical empirical distribution. This optimization process enables the generator network to generate high-fidelity water use behavior sequences from random noise vectors. Specifically, the discriminator network is used as a dynamic, differentiable evaluation function to provide the generator network with continuous, directional gradient guidance regarding the fidelity of its output sequence. Instead of directly fitting the data, the generator network indirectly learns the manifold structure of the real data distribution by minimizing a loss function aimed at deceiving the discriminator. This enables the generator network to capture the complex, nonlinear, and high-dimensional joint probability distribution of historical water use behavior, including temporal correlations between events, statistical regularities in duration, and multi-modal combinations of flow and temperature. The resulting generative capability is achieved by encoding the underlying generative patterns of real water use behavior within the generator network. S4.5. Input a guiding noise vector into the trained generator network.

[0061] Furthermore, the sequence synthesis process is initiated by inputting a guiding noise vector into the trained generator network. There are several ways to construct the guiding noise vector. One common method is to superimpose a pre-defined direction or conditional vector onto the base noise vector sampled from a standard normal distribution. Another method is to construct the guiding noise vector by sampling or interpolating along the direction connecting the normal pattern representation points and the target extreme pattern representation points in the latent space of the generator network. The constructed guiding noise vector is directly input into the trained generator network as the starting seed for the generation process.

[0062] Specifically, the input latent space of a trained generator network possesses a semantic structure, with different directions or regions within the latent space corresponding to different semantic features of the output sequence. By carefully constructing guiding noise vectors, the generation process can be actively intervened and guided, steerd towards the desired semantic direction. For example, superimposing a vector pointing to high-flow semantics on the base noise can guide the generator to produce sequences with an overall high flow level. This guiding input mechanism transforms the generator from a completely random, blind generator into an intelligent generator capable of a certain degree of attribute control as needed. This provides a direct technical means for subsequent steps to purposefully synthesize specific types of extreme conditions, which is the key difference between this method and conventional GAN ​​applications that can only generate random, realistic sequences.

[0063] S4.6 The guiding noise vector points to the region with high intensity and high volatility in the distribution learned by the generator network, driving the generator network to output a time series of virtual user water usage behavior.

[0064] Furthermore, regions corresponding to high intensity and high volatility are identified. This is typically achieved by analyzing the statistical characteristics of sequences in historical user water use behavior datasets and correlating these characteristics with the activation patterns in the latent space of the generator network. For example, the average flow rate and flow rate variance of each sequence in the dataset are obtained, and sequences with both high mean and high variance are identified. Then, using an encoder network or feature inversion techniques, the approximate corresponding regions or direction vectors of these extreme sequences in the latent space of the generator network are found. During synthesis, a guiding noise vector pointing towards or along this region is input into the trained generator network. Based on its learned distribution mapping, the generator network decodes this guiding noise vector into a virtual user water use behavior time series that inherits the statistical regularity of historical data in terms of temporal patterns but significantly exceeds that of ordinary samples in terms of intensity and volatility. Specifically, a controllable correlation was established between the latent spatial semantic direction and the macroscopic features of the output sequence, and this correlation was used to actively explore the edges and tails of the data distribution. The guiding noise vector points to high-intensity, high-fluctuation regions, essentially commanding the generator network within the probability distribution it has learned, representing all possible reasonable water use behaviors. For example, it drives the generator to produce sequences simulating the total flow rate continuously approaching pipeline limits due to multiple water points being simultaneously turned on for extended periods, or sequences simulating drastic flow rate jumps caused by extremely irregular water use habits. This generates extreme test cases that far exceed common patterns in historical data but pose a severe challenge to product reliability.

[0065] S5. Based on the time series of virtual user water usage behavior, drive the operation of a high-fidelity digital twin to calculate the overall energy efficiency value of the water heater and generate an energy efficiency and evaluation report of the water heater under extreme operating conditions.

[0066] S5.1 Load the time series of virtual user water usage behavior into a high-fidelity digital twin, and the high-fidelity digital twin generates high-fidelity digital twin operation process data.

[0067] Furthermore, data points such as time, flow rate, and set temperature from the virtual user's water usage behavior time series are sequentially assigned to the corresponding input ports or input variables of the high-fidelity digital twin, according to chronological order. This drives the high-fidelity digital twin to run step by step from its initial state, based on the input time series. During the operation of the high-fidelity digital twin, the state variables, intermediate variables, and output variables of all computing units within the high-fidelity digital twin are recorded at every moment. This complete set of data, recorded in chronological order and reflecting the dynamic response of the high-fidelity digital twin, constitutes the high-fidelity digital twin's operational process data.

[0068] S5.2 Extract energy consumption and effective heat data from the high-fidelity digital twin's operation process data, and use the energy consumption and effective heat data to calculate the overall energy efficiency value of the water heater.

[0069] Furthermore, specific data required for calculations are extracted from the high-fidelity digital twin's operational data. Energy consumption data is extracted by reading the time series reflecting energy input over the entire operational time range from the high-fidelity digital twin's operational data, such as the instantaneous gas consumption power series of a gas water heater or the instantaneous electrical power series of an electric water heater. Effective heat data is extracted by reading the time series of outlet water temperature, inlet water temperature, and water flow rate during operation from the high-fidelity digital twin's operational data. Using the energy consumption data, the total energy consumption is calculated by numerically integrating the instantaneous power series over the entire operational time. Using the effective heat data, the instantaneous effective heat power at each moment is calculated based on the difference between the outlet and inlet water temperatures and the water flow rate, and then this instantaneous effective heat power series is numerically integrated over the entire operational time.

[0070] The overall energy efficiency value expression for a water heater is: ; in, This refers to the overall energy efficiency value of the water heater. For effective heat, Total energy consumption.

[0071] S5.3 Extract the time series of performance indicators from the high-fidelity digital twin's operation process data, and conduct a comprehensive analysis of the performance indicator time series in conjunction with the overall energy efficiency value of the water heater.

[0072] Furthermore, time series of other performance indicators besides energy consumption and effective heat are extracted from the high-fidelity digital twin's operational data. These performance indicator time series may include temperature change curves of key components, internal pressure fluctuation curves, control valve opening change curves, fan or pump speed curves, and characteristic signals such as potential localized overheating or temperature oscillations. A comprehensive analysis of the performance indicator time series is then conducted in conjunction with the overall energy efficiency value of the water heater. The overall energy efficiency value is used as a core evaluation indicator, and its correlation with the aforementioned multiple performance indicator time series is analyzed and cross-interpreted. The analysis includes observing the state of each performance indicator during high energy efficiency operation; identifying which performance indicators first show abnormalities or reach their limits during low or declining energy efficiency stages; and understanding how the interrelationships between performance indicators ultimately manifest in changes in overall energy efficiency.

[0073] S5.4 Based on the comprehensive analysis results, generate an energy efficiency and evaluation report for the water heater under extreme operating conditions.

[0074] Furthermore, the overall energy efficiency value of the water heater, the analysis conclusions of key performance indicators, the identified abnormal patterns or boundary triggering events, and possible risk warnings are organized into a document containing textual descriptions, data tables, and trend charts. The core content of the water heater's energy efficiency and evaluation report under extreme operating conditions includes: a characteristic description of the applied virtual user water usage behavior time series; the overall energy efficiency value of the water heater obtained under this condition and its comparison with the energy efficiency under standard operating conditions; a summary of the performance of each major performance indicator observed during operation; an assessment of the product's stability, reliability, efficiency characteristics, and potential vulnerabilities under this extreme condition; and the possibility of control optimization suggestions based on the analysis results.

[0075] This embodiment also provides a simulation system for testing the energy efficiency of a water heater, including: a construction module that uses the design drawings and component characteristics of the water heater to construct a parameterized unit library for the water heater; The comparison module, under preset basic test conditions, synchronously drives the simulation process of the water heater parameterized unit library and the physical water heater prototype, and iteratively compares the measured data stream of the physical prototype with the simulation output data stream to form a high-fidelity digital twin. The data acquisition module collects historical user water usage data and constructs a historical user water usage behavior dataset. The training module uses a generative adversarial network to train on a historical user water use behavior dataset, and uses the trained generator network to synthesize virtual user water use behavior time series with extreme water use patterns. The assessment report module, based on the time series of virtual user water usage behavior, drives the operation of a high-fidelity digital twin to calculate the overall energy efficiency value of the water heater and generate an energy efficiency and assessment report of the water heater under extreme operating conditions.

[0076] This embodiment also provides a computer device suitable for a simulation method of water heater energy efficiency testing, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the simulation method of water heater energy efficiency testing as proposed in the above embodiment.

[0077] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0078] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the simulation method for testing the energy efficiency of a water heater as proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0079] In summary, this invention combines digital twin generation with Generative Adversarial Networks (GANs) to achieve efficient and high-precision virtual evaluation of water heater performance under extreme operating conditions. By constructing a parameterized component library and using synchronous testing and iterative comparison, the simulation model parameters are driven to evolve autonomously, generating a high-fidelity digital twin that is highly consistent with a specific physical prototype. This solves the problem that traditional simulation models cannot accurately match the characteristics of individual physical objects due to their generalization, ensuring the accuracy of the virtual test object. By using GANs to train on historical water usage data, extreme water usage behavior sequences containing characteristics such as high intensity and high volatility are synthesized, breaking through the limitations of relying on limited standard operating conditions. This achieves automatic coverage of complex and high-pressure usage scenarios, driving the high-fidelity digital twin to run under extreme conditions, and can calculate its overall energy efficiency under harsh conditions and generate an evaluation report.

[0080] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A simulation method for testing the energy efficiency of a water heater, characterized in that: This includes building a parametric unit library for water heaters using design drawings and component characteristics; Under the preset basic test conditions, the simulation process that drives the water heater parameterized unit library to run synchronously is compared with the physical water heater prototype. The measured data stream of the physical prototype is iteratively compared with the simulation output data stream to form a high-fidelity digital twin. Collect historical user water usage data to construct a historical user water usage behavior dataset; Generative adversarial networks are used to train a dataset of historical user water use behavior, and the trained generator network is used to synthesize time series of virtual user water use behavior with extreme water use patterns. Based on the time series of virtual user water usage behavior, a high-fidelity digital twin is driven to run, calculate the overall energy efficiency value of the water heater, and generate an energy efficiency and evaluation report of the water heater under extreme operating conditions.

2. The simulation method for testing the energy efficiency of a water heater as described in claim 1, characterized in that: The construction of the water heater parameterized unit library includes... By utilizing the design drawings and component characteristics of the water heater, the functional components that make up the water heater are identified, and a mathematical relationship description of the water heater components is established. In the mathematical relationship description of water heater components, quantifiable physical quantities that affect energy efficiency are set as undetermined parameters in the mathematical relationship description of water heater components, and standardized sets of input and output variables are formed, which are then integrated into a water heater parameterized unit library.

3. The simulation method for testing the energy efficiency of a water heater as described in claim 2, characterized in that: The simulation process and the physical water heater prototype include, A set of basic test conditions is preset. The parameterized unit library of the water heater is used to parse and match the instruction sequence of the basic test conditions and input variables. The calculation step size is set to balance the simulation accuracy and speed, thus forming the simulation process. Under the preset basic test conditions, the same basic test condition commands are converted into physical control signals executed by the physical water heater prototype through the measurement and control equipment, thereby driving the physical water heater prototype to operate.

4. The simulation method for testing the energy efficiency of a water heater as described in claim 3, characterized in that: The high-fidelity digital twin includes, Collect actual measured data streams from the operation of a physical water heater prototype; The simulation output data stream generated by the simulation process is obtained by reading the values ​​of virtual monitoring points during the simulation process. The measured data stream of the physical prototype is iteratively compared with the simulation output data stream; Based on the iterative comparison results of the measured data stream of the physical prototype and the simulation output data stream, the undetermined parameters in the parameterized unit library of the water heater are driven to evolve autonomously. When the dynamic response of the simulation process converges with the actual operating response of the physical water heater prototype within the preset tolerance range, the state of the water heater parameterized unit library is recorded to form a high-fidelity digital twin.

5. The simulation method for testing the energy efficiency of a water heater as described in claim 4, characterized in that: The historical user water usage behavior dataset includes, Collect historical user water usage data from smart water heaters and perform cleaning and formatting processes; The historical user water usage data, after being cleaned and formatted, is integrated into a historical user water usage behavior dataset.

6. The simulation method for testing the energy efficiency of a water heater as described in claim 5, characterized in that: The historical user water usage behavior dataset is used for training, including... Time series samples are extracted from historical user water usage behavior datasets, and the time series samples are input into a generative adversarial network for adversarial training to obtain a generator network and a discriminator network. The generator network receives random noise vectors to obtain simulated time series, and the discriminator network receives real time series samples and performs authenticity judgment against the simulated time series generated by the generator network. During adversarial training, the generator network and the discriminator network engage in iterative game and parameter updates using time-series samples as a reference.

7. The simulation method for testing the energy efficiency of a water heater as described in claim 6, characterized in that: The virtual user water usage behavior time series includes, The generator network adjusts its internal parameters based on the feedback signal of the self-discriminator network, so that the simulated time series obtained by the generator network is close to the real time series samples in the historical user water use behavior dataset in terms of statistical characteristics, time patterns and numerical distribution, thereby enabling the generator network to generate water use behavior sequences from random noise. By inputting a guiding noise vector into the trained generator network; The guiding noise vector points to the region of high intensity and high volatility in the distribution learned by the generator network, driving the generator network to output a time series of virtual user water usage behavior.

8. The simulation method for testing the energy efficiency of a water heater as described in claim 7, characterized in that: The calculation of the overall energy efficiency value of the water heater includes... The time series of virtual user water usage behavior is loaded into a high-fidelity digital twin, and the operation of the high-fidelity digital twin generates high-fidelity digital twin operation process data. Energy consumption and effective heat data are extracted from the high-fidelity digital twin's operation data, and the overall energy efficiency value of the water heater is calculated using the energy consumption and effective heat data.

9. The simulation method for testing the energy efficiency of a water heater as described in claim 8, characterized in that: The energy efficiency and assessment report includes, The time series of performance indicators are extracted from the high-fidelity digital twin's operation data, and the time series of performance indicators are comprehensively analyzed in combination with the overall energy efficiency value of the water heater. Based on the comprehensive analysis results, an energy efficiency and evaluation report of the water heater under extreme operating conditions is generated.

10. A simulation system for testing the energy efficiency of a water heater, based on the simulation method for testing the energy efficiency of a water heater according to any one of claims 1 to 9, characterized in that: This includes building modules, which utilize the design drawings and component characteristics of water heaters to construct a parametric unit library for water heaters; The comparison module, under preset basic test conditions, synchronously drives the simulation process of the water heater parameterized unit library and the physical water heater prototype, and iteratively compares the measured data stream of the physical prototype with the simulation output data stream to form a high-fidelity digital twin. The data acquisition module collects historical user water usage data and constructs a historical user water usage behavior dataset. The training module uses a generative adversarial network to train on a historical user water use behavior dataset, and uses the trained generator network to synthesize virtual user water use behavior time series with extreme water use patterns. The assessment report module, based on the time series of virtual user water usage behavior, drives the operation of a high-fidelity digital twin to calculate the overall energy efficiency value of the water heater and generate an energy efficiency and assessment report of the water heater under extreme operating conditions.

Images