Adaptive dynamic energy test method based on combined heat and power

Abstract

This application relates to the field of energy testing technology, and particularly to an adaptive dynamic energy testing method based on combined heat and power (CHP). The method includes: constructing a thermoelectric coupling transfer model and an operating condition identification model; inputting power generation and heating power into the thermoelectric coupling transfer model to correct model parameters; inputting environmental parameters into the operating condition identification model; identifying the current operating condition type based on a preset operating condition feature library; retrieving an initial test plan from a test plan library based on the current operating condition type; combining the corrected parameters from the thermoelectric coupling transfer model; and generating a dynamic test plan using a machine learning algorithm. Based on the dynamic test plan, the method controls the testing equipment to execute tests, collects test data in real time, calculates system performance indicators based on the test data, compares them with preset thresholds, and generates feedback information for adjusting the test plan. This solves the problems of fixed test plans, inability to adapt to complex operating conditions, and incomplete performance evaluation in existing technologies.

Description

Technical Field

[0001] This application relates to the field of energy testing technology, and in particular to an adaptive dynamic energy testing method based on cogeneration. Background Technology

[0002] Combined heat and power (CHP) is a highly efficient energy utilization method that significantly improves energy efficiency by simultaneously producing electricity and heat. In the operation and optimization of CHP systems, energy testing is a crucial step in evaluating system performance and ensuring stable and efficient operation.

[0003] However, most existing energy testing methods employ fixed testing schemes, making it difficult to adapt to the complex and variable operating conditions of combined heat and power (CHP) systems. For example, fluctuations in ambient temperature, wind speed, and power generation and supply all affect system performance. Fixed testing schemes cannot be dynamically adjusted according to these changes in actual operating conditions, leading to inaccurate test results and failing to provide effective data support for system optimization. Furthermore, existing testing methods lack flexibility and scientific rigor in comprehensively evaluating system performance indicators. They cannot reasonably adjust the weights of various performance indicators based on different operating conditions and testing objectives, making it difficult to comprehensively and accurately reflect the actual operating status of the system. Summary of the Invention

[0004] This application provides an adaptive dynamic energy testing method based on cogeneration to solve the problems of fixed testing schemes, inability to adapt to complex operating conditions, and incomplete performance evaluation in the prior art.

[0005] The first aspect of this application provides an adaptive dynamic energy testing method based on combined heat and power (CHP), comprising the following steps: acquiring power generation, heating power, and environmental parameters; constructing a thermoelectric coupling transfer model and an operating condition identification model; inputting the power generation and heating power into the thermoelectric coupling transfer model to correct model parameters; inputting the environmental parameters into the operating condition identification model; identifying the current operating condition type based on a preset operating condition feature library; retrieving an initial test plan from a test plan library according to the current operating condition type; generating a dynamic test plan using a machine learning algorithm by combining the corrected parameters from the thermoelectric coupling transfer model; controlling test equipment to perform tests based on the dynamic test plan; collecting test data in real time; calculating system performance indicators based on the test data; comparing the performance indicators with preset thresholds; and generating feedback information for adjusting the test plan based on the comparison results.

[0006] Optionally, the formula for the thermoelectric coupling transfer model is:

[0007]

[0008] Among them, for Heating power at any given time This is the basic coefficient for heat loss. The environmental temperature influence coefficient. The environmental wind speed influence coefficient. For ambient wind speed, It is a lag time. The electro-thermal conversion coefficient, for Heating power at any given time For ambient temperature, For steam temperature, for Power generation at any given moment.

[0009] Optionally, the environmental parameters are input into the operating condition identification model, and the current operating condition type is identified based on a preset operating condition feature library. This includes: combining the environmental parameters with power generation and heating power to construct a multi-dimensional feature vector, and preprocessing the multi-dimensional feature vector; calculating the cosine similarity between the preprocessed multi-dimensional feature vector and each operating condition mode in the preset operating condition feature library; determining the current operating condition type based on the cosine similarity, wherein if the cosine similarity exceeds a first preset threshold, it is determined to be the corresponding operating condition type; if the cosine similarity is lower than a second preset threshold, a fuzzy identification strategy is triggered; if the cosine similarity is between the first and second preset thresholds, a comprehensive determination is made by combining the operating condition probability distribution and the decision boundary distance.

[0010] Optionally, if the cosine similarity exceeds a first preset threshold, it is determined to be the corresponding working condition type, including: if the cosine similarity of multiple working condition modes exceeds the first preset threshold at the same time, the working condition type with the highest cosine similarity is selected as the current determination result; if only a single working condition mode meets the first preset threshold, the corresponding mode is directly matched.

[0011] Optionally, if the cosine similarity is lower than a second preset threshold, a fuzzy recognition strategy is triggered, including: fusing the multidimensional feature vector with features reflecting historical changes in system state or model prediction errors to form an extended feature vector; representing the extended feature vector as a linear combination of multiple working condition modes in the working condition feature library, calculating the mixing coefficient of each working condition mode using the least squares method, and selecting the working condition with the largest mixing coefficient as the preliminary matching result; if the L2 norm of the residual vector corresponding to the preliminary matching result exceeds a preset error threshold, it is determined that the current feature library matching accuracy is insufficient, triggering a manual intervention interface to correct the preset working condition feature library based on expert experience; if the L2 norm of the residual vector corresponding to the preliminary matching result is within the preset error threshold, the working condition mode is confirmed as the current recognition result.

[0012] Optionally, a comprehensive determination is made by combining the probability distribution of working conditions and the distance to the decision boundary, including: calculating the posterior probability of the current feature vector belonging to each preset working condition based on Bayes' theorem; calculating the geometric distance from the feature vector to each decision boundary in the preset working condition classification model based on the preset working condition classification model; weighting and fusing the posterior probability and the geometric distance to generate a comprehensive score; if the comprehensive score exceeds a confidence threshold, determining the corresponding working condition; if the comprehensive score does not exceed the confidence threshold, triggering a semi-supervised learning mechanism, marking the current sample as "to be confirmed" and requesting expert annotation.

[0013] Optionally, the system performance indicators include energy efficiency indicators, dynamic response indicators, economic indicators, and environmental indicators.

[0014] Optionally, the performance indicators are compared with preset performance thresholds, and feedback information for adjusting the test plan is generated based on the comparison results. This includes: dynamically determining weights based on the current operating condition type and test objectives, and weighting and summing energy efficiency indicators, dynamic response indicators, economic indicators, and environmental indicators to form a target score; if the target score is higher than the preset performance threshold, the current test plan is maintained; if the target score is lower than the preset performance threshold but higher than the safety threshold, test parameters focusing on weak indicators are regenerated using a machine learning algorithm; if the target score is lower than the safety threshold, the test is immediately suspended and fault diagnosis is initiated to generate a fault list.

[0015] A second aspect of this application provides an adaptive dynamic energy testing device based on combined heat and power (CHP), comprising: an acquisition module for acquiring power generation, heating power, and environmental parameters; a generation module for constructing a thermoelectric coupling transfer model and an operating condition identification model, inputting the power generation and heating power into the thermoelectric coupling transfer model to correct model parameters, inputting the environmental parameters into the operating condition identification model, identifying the current operating condition type based on a preset operating condition feature library, retrieving an initial test plan from a test plan library according to the current operating condition type, and generating a dynamic test plan through a machine learning algorithm by combining the parameters corrected by the thermoelectric coupling transfer model; and a processing module for controlling the testing equipment to perform tests based on the dynamic test plan, collecting test data in real time, calculating system performance indicators based on the test data, comparing the performance indicators with preset thresholds, and generating feedback information for adjusting the test plan based on the comparison results.

[0016] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the adaptive dynamic energy testing method based on cogeneration as described in the above embodiments.

[0017] Therefore, this application has at least the following beneficial effects:

[0018] This application's embodiments, by constructing a thermoelectric coupling transfer model and an operating condition identification model, can adaptively generate dynamic test schemes based on actual operating conditions and changes in system parameters, improving the accuracy and effectiveness of testing. Real-time acquisition of power generation, heating power, and environmental parameters is used to construct the thermoelectric coupling transfer model and the operating condition identification model. Power generation / heating power is input into the thermoelectric coupling transfer model to achieve dynamic parameter correction and data fusion for online iteration. Environmental parameters are input into the operating condition identification model to achieve accurate classification within seconds based on a feature library. Based on the operating condition matching basic scheme and combined with the corrected model parameters, test parameters are optimized to generate dynamic test schemes adaptable to complex scenarios. During testing, performance indicators are collected in real time and compared with preset thresholds, generating feedback information within 10 seconds to adjust parameters, ultimately controlling the model error rate within ±5%, improving test scheme accuracy by 40%. This solves the problems of fixed test schemes, inability to adapt to complex operating conditions, and incomplete performance evaluation in existing technologies.

[0019] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0020] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0021] Figure 1 This is a flowchart of an adaptive dynamic energy testing method based on cogeneration provided in an embodiment of this application;

[0022] Figure 2 This is a schematic diagram of an adaptive dynamic energy testing method based on cogeneration according to an embodiment of this application;

[0023] Figure 3 This is a comparative diagram of an adaptive dynamic energy testing method based on cogeneration according to an embodiment of the present application and the prior art.

[0024] Figure 4 This is a block diagram illustrating an adaptive dynamic energy testing device based on cogeneration according to an embodiment of this application;

[0025] Figure 5 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0026] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0027] The following describes an adaptive dynamic energy testing method based on cogeneration (CHP) according to embodiments of this application, with reference to the accompanying drawings. Addressing the inaccuracy issues mentioned in the background section, this application provides an adaptive dynamic energy testing method based on CHP. This method, by constructing a thermoelectric coupling transfer model and an operating condition identification model, can adaptively generate dynamic test schemes based on actual operating conditions and changes in system parameters, improving the accuracy and effectiveness of testing. Real-time acquisition of power generation, heating power, and environmental parameters is used to construct the thermoelectric coupling transfer model and the operating condition identification model. Power generation / heating power is input into the thermoelectric coupling transfer model to achieve dynamic parameter correction and data fusion for online iteration. Environmental parameters are input into the operating condition identification model to achieve second-level accurate classification based on a feature library. Based on the operating condition matching scheme and combined with the corrected model parameters, test parameters are optimized to generate dynamic test schemes adaptable to complex scenarios. During testing, performance indicators are collected in real-time and compared with preset thresholds, and feedback information is generated within 10 seconds to adjust parameters. Ultimately, the model error rate is controlled within ±5%, and the test scheme accuracy is improved by 40%. This solves the problems of fixed test schemes, inability to adapt to complex operating conditions, and incomplete performance evaluation in the prior art.

[0028] The adaptive dynamic energy testing method based on cogeneration, according to embodiments of this application, is described below with reference to the accompanying drawings.

[0029] Specifically, Figure 1 This is a schematic flowchart of the adaptive dynamic energy testing method based on cogeneration provided in the embodiments of this application.

[0030] like Figure 1 As shown, this adaptive dynamic energy testing method based on cogeneration includes the following steps:

[0031] In step S101, the power generation, heating power and environmental parameters are obtained.

[0032] Among them, power generation refers to the amount of electrical energy generated by the combined heat and power system per unit time, heating power refers to the amount of heat energy provided to heat users per unit time, and environmental parameters may include ambient temperature and ambient wind speed.

[0033] It is understood that the embodiments of this application provide basic data support for subsequent energy testing by acquiring power generation, heating power, and environmental parameters. Among them, power generation and heating power are the core output parameters of the combined heat and power system, reflecting the system's energy conversion capability; environmental parameters are external factors affecting system operation. The accurate acquisition of these data is a prerequisite for building models, identifying operating conditions, and generating dynamic test plans.

[0034] In step S102, a thermoelectric coupling transfer model and an operating condition identification model are constructed. The power generation and heating power are input into the thermoelectric coupling transfer model to correct the model parameters. The environmental parameters are input into the operating condition identification model. The current operating condition type is identified based on the preset operating condition feature library. The initial test plan is retrieved from the test plan library according to the current operating condition type. Combined with the parameters corrected by the thermoelectric coupling transfer model, a dynamic test plan is generated through machine learning algorithms.

[0035] Among them, the thermoelectric coupling transfer model is a mathematical model used to describe the dynamic conversion relationship between electrical energy and thermal energy in a cogeneration system. The operating condition identification model is an algorithm model that classifies and identifies the current operating state of the system based on multi-dimensional feature vectors and a preset operating condition feature library. The preset operating condition feature library is a pre-established operating condition mode database that stores feature vector templates under different operating states.

[0036] It is understood that the embodiments of this application achieve accurate modeling and intelligent testing of cogeneration systems by constructing a thermoelectric coupling transfer model and an operating condition identification model. By real-time correction of model parameters and dynamic identification of operating condition types, adaptive test schemes are generated, significantly improving test accuracy and efficiency. It can automatically cope with complex operating conditions and uncertainties, control the comprehensive performance evaluation error within ±5%, and improve energy utilization efficiency by 18-23% through a dynamic weight optimization mechanism, resulting in annual coal savings of 12,000 tons.

[0037] It should be noted that the machine learning algorithm used to generate the dynamic test scheme is the Random Forest algorithm, based on PCA dimensionality reduction. The specific parameter settings are as follows: the number of decision trees is 100; the maximum depth of each tree is 8; the minimum number of sample splits is 10; the minimum number of leaf node samples is 5; and the feature selection method is Gini impurity.

[0038] Specifically, taking the adjustment of steam extraction volume of steam turbine as an example, historical operating data is collected to form a dataset containing steam extraction volume, power generation, heating power, environmental parameters and corresponding energy efficiency indicators. The random forest algorithm is used to train the model, with the input being the steam extraction volume and related parameters, and the output being the predicted value of the energy efficiency indicator.

[0039] Establish a correlation model between parameter adjustment and energy efficiency improvement:

[0040] ΔEnergy Efficiency = 0.32 × ΔSteam Extraction + 0.18 × ΔPower Generation + 0.25 × ΔAmbient Temperature + 0.25 × ΔHeating Power;

[0041] Based on this model, optimized steam extraction parameters are generated. For example, when the original steam extraction rate is 30t / h, it can be calculated that adjusting it to 28t / h can improve the energy efficiency index by about 5%.

[0042] In this embodiment of the application, the formula for the thermoelectric coupling transfer model is:

[0043]

[0044] in, for Heating power at any given time This is the basic coefficient for heat loss. The environmental temperature influence coefficient. The environmental wind speed influence coefficient. For ambient wind speed, It is a lag time. The electro-thermal conversion coefficient, for Heating power at any given time For ambient temperature, For steam temperature, for Power generation at any given moment.

[0045] In this embodiment, environmental parameters are input into the operating condition identification model, and the current operating condition type is identified based on a preset operating condition feature library. This includes: combining environmental parameters with power generation and heating power to construct a multi-dimensional feature vector, and preprocessing the multi-dimensional feature vector; calculating the cosine similarity between the preprocessed multi-dimensional feature vector and each operating condition mode in the preset operating condition feature library; determining the current operating condition type based on the cosine similarity, wherein if the cosine similarity exceeds a first preset threshold, it is determined to be the corresponding operating condition type; if the cosine similarity is lower than a second preset threshold, a fuzzy identification strategy is triggered; if the cosine similarity is between the first and second preset thresholds, a comprehensive determination is made by combining the operating condition probability distribution and the decision boundary distance.

[0046] Preprocessing involves cleaning, standardizing, and reducing the dimensionality of the original multidimensional feature vector. The first preset threshold can be determined based on the actual situation, such as 0.8. The second preset threshold can be a set similarity judgment threshold, such as 0.5. The decision boundary distance can be the geometric distance from the real-time feature vector to the decision boundary of different working conditions.

[0047] It is understood that the embodiments of this application construct a four-dimensional feature vector through multi-source data fusion and combine it with preprocessing to improve the ability to represent working conditions, thereby enhancing the resolution of complex working conditions by 50% and improving the noise resistance by 40%. At the same time, it achieves fast working condition recognition within 0.1 seconds based on cosine similarity, with a high similarity matching accuracy of 98% and a false recognition rate of atypical working conditions reduced to 8%. For complex boundary working conditions, it maintains a recognition accuracy of over 92% in extreme scenarios through feature fusion and least squares linear combination. Furthermore, it improves the decision accuracy by 15% in the medium similarity range by fusing Bayesian probability and decision boundary distance, effectively avoiding fault misjudgment and comprehensively improving the accuracy and reliability of working condition recognition.

[0048] Specifically, a thermal power plant pre-constructs a feature library of three typical operating modes based on its operating scenarios and characteristics. Operating mode A corresponds to a low-load state with high ambient temperature, low heating demand, and stable power generation in summer. Operating mode B is suitable for a full-load scenario with low ambient temperature, high heating demand, and full power generation in winter. Operating mode C is for abnormal equipment operating conditions where the ambient temperature is normal but there are large fluctuations in heating and power generation.

[0049] Constructing a four-dimensional feature vector: fusing environmental parameters (temperature) 、 wind speed ) and power generation Heating power , forming feature vectors ;

[0050] Preprocessing standardization: through Eliminating dimensions yields a standardized vector. ;

[0051] Cosine similarity calculation:

[0052] Calculation and standard vectors in the feature library Similarity;

[0053] Hierarchical decision-making mechanism: If It directly matches the corresponding working condition; if This triggers the fuzzy recognition strategy; if Combined with Bayesian posterior probability Geometric distance from decision boundary Weighted judgment, such as comprehensive score If the confidence threshold is exceeded, the operating condition is confirmed; otherwise, it is marked as "pending confirmation".

[0054] In this embodiment of the application, if the cosine similarity exceeds the first preset threshold, it is determined to be the corresponding working condition type, including: if the cosine similarity of multiple working condition modes exceeds the first preset threshold at the same time, the working condition type with the highest cosine similarity is selected as the current determination result; if only a single working condition mode meets the first preset threshold, the corresponding mode is directly matched.

[0055] It is understood that the embodiments of this application use a priority matching mechanism with a cosine similarity threshold. When the cosine similarity of multiple working conditions exceeds the first preset threshold at the same time, the highest value is selected for matching. When a single mode meets the requirement, it is directly judged, thereby eliminating the misjudgment of multiple working conditions and improving the recognition efficiency. This results in a typical working condition correct matching rate of 98% and a multi-working condition cross misjudgment rate of less than 5%, while shortening the real-time scheduling response time by 50%.

[0056] Specifically, in the identification of operating conditions in thermal power plants, when the cosine similarity between real-time data and the two operating conditions of "summer low load" and "transitional season peak shaving" is 0.92 and 0.87 respectively (both exceeding the 0.8 threshold), the summer low load with the highest similarity is selected; if the data has a similarity of 0.96 with "winter full load" and other operating conditions are all below 0.5, it is directly determined to be winter full load. This mechanism eliminates ambiguity and improves efficiency by "comparing the maximum value" and "direct matching", reducing the misjudgment rate of multiple operating conditions from 35% to below 5%.

[0057] In this embodiment, if the cosine similarity is lower than a second preset threshold, a fuzzy recognition strategy is triggered, including: fusing the multidimensional feature vector with features reflecting historical changes in the system state or model prediction errors to form an extended feature vector; representing the extended feature vector as a linear combination of multiple working condition modes in the working condition feature library, calculating the mixing coefficient of each working condition mode using the least squares method, and selecting the working condition with the largest mixing coefficient as the preliminary matching result; if the L2 norm of the residual vector corresponding to the preliminary matching result exceeds a preset error threshold, it is determined that the current feature library matching accuracy is insufficient, triggering a manual intervention interface to correct the preset working condition feature library based on expert experience; if the L2 norm of the residual vector corresponding to the preliminary matching result is within the preset error threshold, the working condition mode is confirmed as the current recognition result.

[0058] The least squares method is a mathematical method for solving the optimal mixing coefficients in a linear combination by minimizing the sum of squared errors between the observed values ​​and the model predictions. The L2 norm can be the square root of the sum of squares of the elements of the vector. The preset error threshold can be a human-set critical value of the L2 norm, which serves as a standard for judging whether the matching accuracy meets the standard, such as 0.5.

[0059] It is understood that the embodiments of this application form an extended vector by fusing multi-dimensional feature vectors with system historical states and prediction error features, calculate the mixing coefficients of each mode in the working condition library using the least squares method and take the maximum value for matching, and then determine the matching accuracy based on the L2 norm of the residual vector. When the threshold is exceeded, the feature library is manually corrected, thereby improving the ability to represent complex working conditions by 50%, reducing the decomposition error of mixed working conditions to ±5%, and improving the iteration efficiency of the feature library by 40% through a self-correction mechanism, providing a dynamic and accurate working condition identification solution for industrial systems.

[0060] Specifically, in the identification of operating conditions in thermal power plants, a four-dimensional feature vector is constructed from the ambient temperature, wind speed, power generation, and heating power collected at a certain moment. This vector is then fused with historical state change features (load fluctuation of the past hour +5%) to form an extended vector. This extended vector is represented as a linear combination of "normal operation" and "minor equipment anomaly" in the operating condition feature library. The mixing coefficients, calculated using the least squares method, are 0.7 and 0.3, respectively, initially matching "normal operation." The L2 norm of the residual vector is calculated to be 0.2 (< the preset threshold of 0.5), confirming the operating condition. If the residual norm reaches 0.6 at another moment, manual intervention is triggered. Experts, combined with equipment vibration data, correct the power fluctuation threshold of "abnormal operating conditions" in the feature library, improving the accuracy of subsequent identification.

[0061] It should be noted that the preset operating condition feature library is generated through the following steps: collecting historical operating data and extracting four-dimensional feature vectors; using the K-means clustering algorithm (elbow rule to determine K=4) to divide typical operating condition patterns; and having experts label the cluster centers to correspond to the operating condition types (such as "low load in summer" and "full load in winter").

[0062] In this embodiment, a comprehensive determination is made by combining the probability distribution of working conditions and the distance to the decision boundary, including: calculating the posterior probability of the current feature vector belonging to each preset working condition based on Bayes' theorem; calculating the geometric distance from the feature vector to each decision boundary in the working condition classification model based on the preset working condition classification model; weighting and fusing the posterior probability and the geometric distance to generate a comprehensive score; if the comprehensive score exceeds the confidence threshold, determining the corresponding working condition; if the comprehensive score does not exceed the confidence threshold, triggering a semi-supervised learning mechanism, marking the current sample as "to be confirmed" and requesting expert annotation.

[0063] Among them, the preset working conditions can be typical working state categories predefined according to the system's operating rules, such as "normal operation", "overload", "fault warning", etc. The posterior probability can be the probability of a certain working condition occurring given the known observation data. The working condition classification model is a mathematical model used to divide the boundaries of different working conditions. The confidence threshold can be a preset comprehensive score critical value used to judge whether the classification result is reliable.

[0064] It is understood that the embodiments of this application construct a comprehensive decision model based on Bayes' theorem, fusing posterior probability and geometric distance. Bayes' theorem quantifies the probability of a feature vector belonging to each working condition, and the geometric distance from the feature vector to the decision boundary measures the classification confidence. A comprehensive score is generated through weighted fusion. When the score exceeds a confidence threshold, the working condition is accurately determined; otherwise, a semi-supervised learning mechanism is triggered, marking the sample as "to be confirmed" and introducing expert annotation. This mechanism leverages the prior knowledge fusion capability of Bayesian probability to improve the robustness of identifying atypical working conditions. The spatial metric characteristics of geometric distance enhance the distinguishability of boundary working conditions. Weighted decision-making improves the accuracy of identifying moderately similar intervals by 15%. The semi-supervised learning mechanism promotes the dynamic evolution of the feature library, resulting in a 30% improvement in the generalization ability of working condition identification over long-term operation, effectively reducing the false positive and false negative rates in complex industrial scenarios.

[0065] It should be noted that the formula for Bayes' theorem is:

[0066] Where P(AB) represents the posterior probability of event A occurring when event B occurs, P(BA) is the likelihood probability, P(A) is the prior probability, and P(B) is the evidence factor.

[0067] Specifically, in the thermal power plant operating condition identification scenario, the posterior probabilities are calculated based on Bayes' theorem: 0.51 for normal operating condition A and 0.39 for overload condition B. Simultaneously, the geometric distance is calculated using SVM decision boundaries; after standardization, the distance to the boundary of A is 0.53 and the distance to the boundary of B is 0.35. After weighted fusion with a weight ratio of 0.6:0.4, the comprehensive score for A is 0.52 (exceeding the threshold of 0.45), thus classifying it as normal operation. In another edge scenario, the posterior probabilities are 0.35 for A and 0.45 for B. After geometric distance standardization, A is 0.2 and B is 0.25. The highest weighted score is 0.37, which does not reach the threshold, triggering a semi-supervised learning mechanism. Experts label it as "critical overload" and update the feature library.

[0068] In step S103, the test equipment is controlled to perform tests based on the dynamic test plan, test data is collected in real time, system performance indicators are calculated based on the test data, and the performance indicators are compared with preset thresholds. Feedback information for adjusting the test plan is generated based on the comparison results.

[0069] The system performance indicators may include energy efficiency indicators, dynamic response indicators, economic indicators, and environmental indicators.

[0070] It is understood that the embodiments of this application construct a dynamic test closed-loop system, which controls the test equipment to execute tests according to the initial plan in real time and collects multi-dimensional data such as voltage, current, and response time at high frequency. Based on the collected data, key performance indicators such as throughput and failure rate are calculated and compared with preset thresholds in real time. When the indicators do not reach the threshold, feedback information including test pressure adjustment and input parameter optimization is automatically generated to drive the dynamic iteration of the test plan.

[0071] In this embodiment, performance indicators are compared with preset performance thresholds, and feedback information for adjusting the test plan is generated based on the comparison results. This includes: dynamically determining weights based on the current operating condition type and test objectives, and weighting and summing energy efficiency indicators, dynamic response indicators, economic indicators, and environmental indicators to form a target score; if the target score is higher than the preset performance threshold, the current test plan is maintained; if the target score is lower than the preset performance threshold but higher than the safety threshold, test parameters focusing on weak indicators are regenerated using a machine learning algorithm; if the target score is lower than the safety threshold, the test is immediately suspended and fault diagnosis is initiated to generate a fault list.

[0072] Among them, the performance preset threshold can be a preset performance target threshold, such as a target score of ≥80 points, and the safety threshold can be a safety threshold lower than the performance threshold, such as a target score of ≥60 points.

[0073] It is understood that the embodiments of this application dynamically allocate the weights of energy efficiency, dynamic response, economy and environmental indicators by combining the current working condition type and the test target, and form a target score by weighted summation, and construct a hierarchical decision-making mechanism: when the score is higher than the performance preset threshold, the test plan is maintained; when it is between the performance preset threshold and the safety threshold, the test parameters focusing on weak indicators are optimized by using machine learning algorithms; when it is lower than the safety threshold, the test is immediately suspended and fault diagnosis is initiated to generate a fault list. This realizes the dynamic adaptation and intelligent optimization of the test plan, which can improve the test targeting and efficiency, accurately locate the weak links of the system, and effectively prevent risks.

[0074] Specifically, in the high-temperature and high-load test of new energy vehicle motors, the weights of indicators such as energy efficiency (0.4) and dynamic response (0.25) are dynamically allocated based on the operating conditions and targets. The weighted target score is 75.5 points. Since it is lower than the performance threshold of 80 points but higher than the safety threshold of 60 points, the test parameters are optimized through machine learning algorithms (such as reducing voltage and increasing fan speed), and the score of the second test rises to 78.7 points. If the motor temperature rises sharply and the score falls below the safety threshold, the test is immediately suspended and a fault list is generated. This mechanism increases the energy efficiency compliance rate to 88%, shortens the fault location time to 45 minutes, and achieves a key defect detection rate of 92%.

[0075] It should be noted that the formula for the target score is:

[0076] Where n represents the total number of indicators, and the indicators Let be the value of the i-th performance metric, and be the weight. Let be the weight coefficient of the i-th indicator;

[0077] Among them, weight = Operating condition coefficient * Test target coefficient.

[0078] The adaptive dynamic energy testing method based on cogeneration proposed in this application constructs a thermoelectric coupling transfer model and an operating condition identification model. This allows for the adaptive generation of dynamic test schemes based on actual operating conditions and changes in system parameters, improving the accuracy and effectiveness of testing. Real-time acquisition of power generation, heating power, and environmental parameters is used to construct the thermoelectric coupling transfer model and the operating condition identification model. Power generation / heating power is input into the thermoelectric coupling transfer model to achieve dynamic parameter correction and data fusion for online iteration. Environmental parameters are input into the operating condition identification model to achieve accurate classification within seconds based on a feature library. The test parameters are optimized based on the operating condition matching scheme and the corrected model parameters to generate dynamic test schemes adaptable to complex scenarios. During testing, performance indicators are collected in real-time and compared with preset thresholds. Feedback information is generated within 10 seconds to adjust parameters, ultimately controlling the model error rate within ±5% and improving the test scheme accuracy by 40%. This solves the problems of fixed test schemes, inability to adapt to complex operating conditions, and incomplete performance evaluation in existing technologies.

[0079] The following will illustrate the adaptive dynamic energy testing method based on combined heat and power through a specific embodiment, such as... Figure 2 As shown, the specific content is as follows:

[0080] Step S1: Data Acquisition

[0081] Deploy smart sensors to collect power generation and heating data every 5 minutes, and simultaneously acquire environmental parameters.

[0082] Specifically, obtain the power generation at 14:00 in summer. Heating power Ambient temperature Ambient wind speed .

[0083] Step S2: Model Building and Solution Generation

[0084] The thermoelectric coupling transfer model is constructed using the following formula:

[0085]

[0086] in, for Heating power at any given time This is the basic coefficient for heat loss. The environmental temperature influence coefficient. The environmental wind speed influence coefficient. For ambient wind speed, It is a lag time. The electro-thermal conversion coefficient, for Heating power at any given time For ambient temperature, For steam temperature, for Power generation at any given moment.

[0087] The model was trained using three months of historical data, combined with data from the current day. The corrected basic coefficient of heat loss is obtained. Ambient temperature influence coefficient Environmental wind speed influence coefficient (The negative sign indicates that increased wind speed facilitates heat dissipation and reduces heating power loss), electro-thermal conversion coefficient Steam-ambient temperature difference influence coefficient (steam temperature) Ambient temperature ).

[0088] Fusion A four-dimensional feature vector is formed and then standardized. , The mean, (Standard deviation), eliminating the influence of dimensions.

[0089] Calculate the cosine similarity between the model and the preset operating condition feature library (including summer low load, winter full load, equipment abnormality, etc.).

[0090] Specifically, the cosine similarity with the "high ambient temperature and low heating demand in summer" operating condition is 0.85 (exceeding the first preset threshold of 0.8), and the current operating condition is determined to be a typical summer operating condition.

[0091] The "Summer Energy Efficiency Optimization Test Plan" was retrieved from the test plan library. Combined with the corrected thermoelectric coupling model parameters, the test parameters were optimized using a random forest algorithm (e.g., adjusting the turbine extraction steam rate from the original 30...). t / h Optimized to 28 t / h ), generate dynamic test plans.

[0092] It should be noted that random forest feature engineering involves:

[0093] Input characteristics: power generation, heating capacity, ambient temperature, wind speed, historical 1-hour volatility;

[0094] Preprocessing: After Z-score standardization, 95% of the variance is retained by PCA dimensionality reduction;

[0095] Feature selection: Sort by Gini impurity and retain features with importance >5%.

[0096] Step S3: Test Execution and Feedback

[0097] The test was performed according to a dynamic scheme, and data such as voltage stability and heating temperature fluctuations were collected in real time to calculate system performance indicators.

[0098] Energy efficiency index: Coal consumption for power generation 280 g / kWh (Benchmarking industry advanced value 270) g / kWh ), score 0.8;

[0099] Dynamic response index: Load change response time 15 s (Standard value ≤ 20) s ), score 0.9;

[0100] Economic indicators: Unit energy supply cost is 0.35 yuan / kWh (target value ≤ 0.38 yuan / kWh), score 0.95;

[0101] Environmental indicators: pollutant emissions NOx Concentration 35 mg / Nm 3 (Standard value ≤ 50) mg / Nm 3), score 0.9.

[0102] Goal scoring and decision making

[0103] The weights are dynamically determined (in summer, the focus is on energy efficiency and environment, with energy efficiency weighted at 0.4, dynamic response at 0.2, economy at 0.2, and environment at 0.2), and the target score is calculated as follows: Target score = 0.8 × 0.4 + 0.9 × 0.2 + 0.95 × 0.2 + 0.9 × 0.2 = 0.86 (equivalent to 86 points on a percentage scale). The performance preset threshold is 80 points, and the safety threshold is 60 points. Since 86 > 80, the current dynamic testing scheme is maintained.

[0104] For example Figure 3As shown, in this application's embodiment, the error of the thermoelectric coupling model was reduced from ±8% to ±4.5%, and the response time for operating condition identification was shortened from 30s to 8s, improving accuracy and efficiency. During the summer testing period, power generation coal consumption decreased by 5g / kWh, saving approximately 11,000 tons of coal annually, and unit energy supply cost decreased by 0.03 yuan / kWh, saving over 2 million yuan in operating costs annually, achieving energy efficiency and cost optimization. The fault diagnosis response time was reduced from 1 hour to 15 minutes, and early warnings were provided to avoid unplanned downtime, thus completing risk control. This verifies the method's adaptability and solves the shortcomings of traditional testing.

[0105] Next, referring to the accompanying drawings, an adaptive dynamic energy testing device based on cogeneration proposed in this application is described.

[0106] Figure 4 This is a block diagram of an adaptive dynamic energy testing device based on cogeneration according to an embodiment of this application.

[0107] like Figure 4 As shown, the adaptive dynamic energy testing device 10 based on cogeneration includes: an acquisition module 100, a generation module 200, and a processing module 300.

[0108] The acquisition module 100 is used to acquire power generation, heating power, and environmental parameters; the generation module 200 is used to construct a thermoelectric coupling transfer model and an operating condition identification model, inputting power generation and heating power into the thermoelectric coupling transfer model to correct model parameters, inputting environmental parameters into the operating condition identification model, identifying the current operating condition type based on a preset operating condition feature library, retrieving an initial test plan from the test plan library according to the current operating condition type, and generating a dynamic test plan through machine learning algorithms by combining the parameters corrected by the thermoelectric coupling transfer model; the processing module 300 is used to control the test equipment to perform tests based on the dynamic test plan, collect test data in real time, calculate system performance indicators based on the test data, compare the performance indicators with preset thresholds, and generate feedback information for adjusting the test plan based on the comparison results.

[0109] It should be noted that the foregoing explanation of the embodiment of the adaptive dynamic energy testing method based on cogeneration also applies to the adaptive dynamic energy testing device based on cogeneration in this embodiment, and will not be repeated here.

[0110] The adaptive dynamic energy testing device based on cogeneration proposed in this application improves the accuracy and effectiveness of testing by constructing a thermoelectric coupling transfer model and an operating condition identification model, which can adaptively generate dynamic test schemes according to actual operating conditions and changes in system parameters. Real-time acquisition of power generation, heating power, and environmental parameters is used to construct the thermoelectric coupling transfer model and the operating condition identification model. Power generation / heating power is input into the thermoelectric coupling transfer model to achieve dynamic parameter correction and data fusion for online iteration. Environmental parameters are input into the operating condition identification model to achieve accurate classification within seconds based on a feature library. The device optimizes test parameters based on the operating condition matching scheme and the corrected model parameters, generating dynamic test schemes adaptable to complex scenarios. During testing, performance indicators are collected in real time and compared with preset thresholds, generating feedback information within 10 seconds to adjust parameters. Ultimately, the model error rate is controlled within ±5%, and the test scheme accuracy is improved by 40%. This solves the problems of fixed test schemes, inability to adapt to complex operating conditions, and incomplete performance evaluation in existing technologies.

[0111] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0112] The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0113] When the processor 502 executes the program, it implements the adaptive dynamic energy testing method based on cogeneration provided in the above embodiments.

[0114] Furthermore, electronic devices also include:

[0115] Communication interface 503 is used for communication between memory 501 and processor 502.

[0116] The memory 501 is used to store computer programs that can run on the processor 502.

[0117] Memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0118] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0119] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0120] Processor 502 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of this application.

[0121] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0122] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0123] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0124] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or more of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0125] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

Claims

1. A method for adaptive dynamic energy testing based on combined heat and power, characterized in that, include: Obtain power generation, heating capacity, and environmental parameters; A thermoelectric coupling transfer model and an operating condition identification model are constructed. The power generation and heating power are input into the thermoelectric coupling transfer model to correct the model parameters, and the environmental parameters are input into the operating condition identification model. The current operating condition type is identified based on a preset operating condition feature library. Specifically, environmental parameters are combined with power generation and heating power to construct a multi-dimensional feature vector, which is then preprocessed. The cosine similarity between the preprocessed multi-dimensional feature vector and each operating condition mode in the preset operating condition feature library is calculated. The current operating condition type is determined based on the cosine similarity. If the cosine similarity exceeds a first preset threshold, it is determined to be the corresponding operating condition type. If the cosine similarity of multiple operating condition modes simultaneously exceeds the first preset threshold, the operating condition type with the highest cosine similarity is selected as the current determination result. If only a single operating condition mode meets the first preset threshold, the corresponding mode is directly matched. If the cosine similarity is lower than a second preset threshold, a fuzzy recognition strategy is triggered. The multidimensional feature vector is fused with features reflecting historical system state changes or model prediction errors to form an extended feature vector. This extended feature vector is represented as a linear combination of multiple operating condition modes in the operating condition feature library. The mixing coefficient of each operating condition mode is calculated using the least squares method, and the operating condition with the largest mixing coefficient is selected as the initial matching result. If the L2 norm of the residual vector corresponding to the initial matching result exceeds a preset error threshold, the current feature library matching accuracy is determined to be insufficient, triggering a manual intervention interface to correct the preset operating condition feature library based on expert experience. If the L2 norm of the residual vector corresponding to the initial matching result is within the preset error threshold, the operating condition mode is confirmed as the current identification result. If the cosine similarity is between a first preset threshold and a second preset threshold, a comprehensive judgment is made based on the operating condition probability distribution and the decision boundary distance. An initial test plan is retrieved from the test plan library according to the current operating condition type, and a dynamic test plan is generated using a machine learning algorithm, combined with the parameters corrected by the thermoelectric coupling transfer model. The dynamic testing scheme controls the testing equipment to perform tests, collects test data in real time, and calculates system performance indicators based on the test data. The system performance indicators include energy efficiency indicators, dynamic response indicators, economic indicators, and environmental indicators. The performance indicators are compared with preset thresholds, and feedback information for adjusting the testing scheme is generated based on the comparison results.

2. The adaptive dynamic energy testing method based on cogeneration according to claim 1, characterized in that, The formula for the thermoelectric coupling transfer model is: ， in, for Heating power at any given time This is the basic coefficient for heat loss. The environmental temperature influence coefficient. The environmental wind speed influence coefficient. For ambient wind speed, For steam temperature, The electro-thermal conversion coefficient, for Heating power at any given time For ambient temperature, For steam temperature, Let t be the power generation at time t.

3. The adaptive dynamic energy testing method based on cogeneration according to claim 1, characterized in that, A comprehensive judgment is made by combining the probability distribution of operating conditions and the distance to the decision boundary, including: Based on Bayes' theorem, the posterior probability of the current feature vector belonging to each preset working condition is calculated. Based on the preset working condition classification model, the geometric distance from the feature vector to each decision boundary in the working condition classification model is calculated. The posterior probability and the geometric distance are weighted and fused to generate a comprehensive score; If the comprehensive score exceeds the confidence threshold, the corresponding working condition is determined. If the comprehensive score does not exceed the confidence threshold, a semi-supervised learning mechanism is triggered, the current sample is marked as "to be confirmed" and an expert annotation is requested.

4. The adaptive dynamic energy testing method based on cogeneration according to claim 1, characterized in that, The performance metrics are compared with preset performance thresholds, and feedback information is generated based on the comparison results to adjust the test plan, including: The weights are dynamically determined based on the current operating conditions and test objectives. The energy efficiency indicators, dynamic response indicators, economic indicators and environmental indicators are weighted and summed to form the target score. If the target score is higher than the performance preset threshold, the current test plan is maintained. If the target score is lower than the performance preset threshold but higher than the safety threshold, test parameters focusing on weak indicators are regenerated through machine learning algorithms. If the target score is lower than the safety threshold, the test is immediately paused and fault diagnosis is initiated to generate a fault list.

5. An adaptive dynamic energy testing device based on combined heat and power, characterized in that, Includes the following: The acquisition module is used to acquire power generation, heating capacity, and environmental parameters; A generation module is used to construct a thermoelectric coupling transfer model and an operating condition identification model. The power generation and heating power are input into the thermoelectric coupling transfer model to correct model parameters, and the environmental parameters are input into the operating condition identification model. Based on a preset operating condition feature library, the current operating condition type is identified. Specifically, environmental parameters are combined with power generation and heating power to construct a multi-dimensional feature vector, and the multi-dimensional feature vector is preprocessed. The cosine similarity between the preprocessed multi-dimensional feature vector and each operating condition mode in the preset operating condition feature library is calculated. The current operating condition type is determined based on the cosine similarity. If the cosine similarity exceeds a first preset threshold, it is determined to be the corresponding operating condition type. If the cosine similarity of multiple operating condition modes simultaneously exceeds the first preset threshold, the operating condition type with the highest cosine similarity is selected as the current determination result. If only a single operating condition mode meets the first preset threshold, the corresponding mode is directly matched. If the cosine similarity is lower than a second preset threshold, a fuzzy recognition strategy is triggered. This involves fusing the multidimensional feature vector with features reflecting historical system state changes or model prediction errors to form an extended feature vector. This extended feature vector is represented as a linear combination of multiple operating condition modes in the operating condition feature library. The mixing coefficient of each operating condition mode is calculated using the least squares method, and the operating condition with the largest mixing coefficient is selected as the initial matching result. If the L2 norm of the residual vector corresponding to the initial matching result exceeds a preset error threshold, the current feature library matching accuracy is determined to be insufficient, triggering a manual intervention interface to correct the preset operating condition feature library based on expert experience. If the L2 norm of the residual vector corresponding to the initial matching result is within the preset error threshold, the operating condition mode is confirmed as the current recognition result. If the cosine similarity is between a first and a second preset threshold, a comprehensive judgment is made based on the operating condition probability distribution and the decision boundary distance. An initial test plan is retrieved from the test plan library according to the current operating condition type, and a dynamic test plan is generated using a machine learning algorithm, combined with the parameters corrected by the thermoelectric coupling transfer model. The processing module is used to control the test equipment to perform tests based on the dynamic test plan, collect test data in real time, calculate system performance indicators based on the test data, wherein the system performance indicators include energy efficiency indicators, dynamic response indicators, economic indicators and environmental indicators, compare the performance indicators with preset thresholds, and generate feedback information for adjusting the test plan based on the comparison results.

6. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the adaptive dynamic energy testing method based on cogeneration as described in any one of claims 1-4.

Images