Frequency conversion speed regulation energy-saving control method and system based on load dynamic identification

Abstract

The invention relates to the technical field of variable frequency regulation, in particular to a variable frequency speed regulation energy-saving control method and system based on load dynamic identification, and the method comprises the steps: constructing a multi-dimensional time sequence structure for data collected in a current operation period of a system, and identifying the dynamic trend of a current load based on the sequence; establishing a candidate frequency set, outputting a power prediction value corresponding to the candidate frequency set through a power estimation model, calculating a unit energy-saving potential after the current frequency is adjusted to the candidate frequency through an energy-saving potential algorithm, and inputting the candidate frequency corresponding to the unit energy-saving potential greater than a threshold value into a feasible frequency set; on the basis of the feasible frequency set and the corresponding unit energy-saving potential, establishing a comprehensive scoring function, and selecting an optimal target frequency through the comprehensive scoring function; and inputting the optimal target frequency and the current operation frequency into a feedback correction function, outputting an execution frequency of the frequency converter through the feedback correction function, and converting the execution frequency into a corresponding control instruction.

Description

Technical Field

[0001] This invention relates to the field of frequency conversion regulation technology, and more particularly to a frequency conversion speed regulation energy-saving control method and system based on dynamic load identification. Background Technology

[0002] In energy-saving control of motor-type loads, variable frequency speed regulation has become the mainstream method. However, in engineering sites, adjustment methods based on threshold triggering or fixed rules are still commonly used, relying more on instantaneous quantities or empirical curves, which makes it difficult to characterize the dynamic characteristics of load rise, fall, and fluctuation in a short period of time. Common problems with this approach include: First, it only looks at the current state and does not identify the direction and speed of the upcoming change, resulting in inaccurate timing of frequency adjustment; second, it treats frequency adjustment as "zero cost," failing to include the cost of frequency change itself and equipment inertia in the calculation, which easily leads to frequent, small-step, and unconstrained adjustments, resulting in system fluctuations, mechanical shocks, and energy losses; third, energy efficiency evaluation is independent of the control chain, often based on rough statistical averages or post-event summaries, making it difficult to form quantitative indicators that are useful and directly usable for the next cycle; fourth, at the execution level, it often uses the method of directly writing the target frequency, ignoring the impact of acceleration and deceleration paths on motor current, pipeline hydraulic or pneumatic system stability under different load scenarios; fifth, data sources are scattered and loosely organized in time, and historical windows are not used in a structured way, failing to form a reusable expression of operating status. As a result, the system is prone to problems such as unstable energy-saving effects, mismatch between adjustment actions and physical constraints, uninterpretable strategies, and difficulty in convergence during long-term operation under complex operating conditions, which restricts further improvement in variable frequency energy saving. To address these pain points, a systematic approach is needed that is based on real-time status, uses trend recognition as the entry point, measures both energy-saving benefits and physical costs as the core, and provides executable paths as the exit point. This approach would sequentially connect identification, evaluation, decision-making, and execution in an engineering manner, achieving stable and reusable energy-saving control without altering the existing hardware infrastructure. Summary of the Invention

[0003] The purpose of this invention is to provide a variable frequency speed regulation energy-saving control method and system based on dynamic load identification, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] The variable frequency speed regulation energy-saving control method based on dynamic load identification includes:

[0006] A multidimensional time series is constructed from the collected data during the current system cycle. The multidimensional time series is then input into a combined structure of convolutional layers and attention weighting modules, and trend labels are output through the combined structure.

[0007] A candidate frequency set is established, and the power prediction value corresponding to the candidate frequency set is output through the power estimation model. Based on the power prediction value, the unit energy saving potential after adjusting from the current frequency to the candidate frequency is calculated through the energy saving potential algorithm. The candidate frequencies corresponding to the unit energy saving potential greater than the threshold are input into the feasible frequency set. The energy saving potential algorithm includes the estimated value of the average power in the most recent period at the current frequency, the power prediction value of the candidate frequency, and a trend weighting term. The trend weighting term is used to automatically adjust the sensitivity of the energy saving estimation according to different trend states.

[0008] Based on the set of feasible frequencies and their corresponding unit energy-saving potential, a comprehensive scoring function is established according to the dual constraints of energy-saving efficiency and operational safety. The set of feasible frequencies is scored by the comprehensive scoring function to generate a scoring sequence, and the optimal target frequency is selected from multiple feasible frequencies.

[0009] The optimal target frequency and the current operating frequency are input into the feedback correction function, the execution frequency of the frequency converter is output through the feedback correction function, and the execution frequency is converted into the corresponding control command. The feedback correction function includes the weighted result of the optimal target frequency and the current operating frequency and the frequency adjustment execution weight factor.

[0010] The system collects the actual operating power corresponding to the execution frequency and updates the energy-saving assessment indicators, providing dynamic feedback data for trend identification and energy-saving potential prediction in the next cycle.

[0011] Preferably, the step of converting the execution frequency into corresponding control commands specifically includes:

[0012] If the system uses analog control, the execution frequency is mapped to analog voltage or current through proportional conversion;

[0013] If digital communication is used, the execution frequency is converted into a setting register value and written to the inverter control address.

[0014] Preferably, the comprehensive scoring function includes unit energy saving potential, a first penalty factor, a second penalty factor, and the rate of change of the current average power. The first penalty factor is used to control the impact of frequency adjustment amplitude on the score, and the second penalty factor is used to control the risk caused by load response inertia.

[0015] Preferably, the frequency modulation execution weighting factor is used to map the system state to the frequency modulation speed, balancing the conflict between regulation response and stability. The frequency modulation execution weighting factor includes the comprehensive score corresponding to the optimal target frequency, the smallest positive number to prevent division by zero, the power change rate of the current load, the trend label, the empirical adjustment coefficient, and the weight correction coefficient.

[0016] Preferably, the energy-saving assessment index represents the proportion of energy-saving effect brought about by the current frequency regulation strategy. The energy-saving assessment index is generated by an assessment function, which includes a reference power, a very small positive number to prevent the denominator from being zero, and a power assessment index.

[0017] Preferably, the power evaluation index is generated by combining the instantaneous power value with the frequency modulation execution weighting factor, and the instantaneous power value is collected by a three-phase power sensor installed at the motor output.

[0018] Preferably, the power estimation model is characterized by being composed of a linear regression structure superimposed with trend factor weighting terms.

[0019] Preferably, the trend labels include load decreasing, load stabilizing, and load increasing.

[0020] Preferably, the collected data includes the current operating frequency of the inverter, the motor current, and the real-time active power. The current operating frequency is obtained through the inverter's communication register. The motor current is collected by a current transformer installed on the motor power supply side, and after signal conditioning, it is input to the controller. The real-time active power can be obtained through the inverter's internal parameters or an independent power monitoring module.

[0021] A variable frequency speed regulation energy-saving control system based on dynamic load identification includes:

[0022] The trend recognition module is used to construct a multidimensional time series from the data collected during the current operating cycle of the system. The multidimensional time series is input into a combined structure of convolutional layer and attention weighting module, and trend labels are output through the combined structure.

[0023] The energy-saving potential assessment module is used to establish a candidate frequency set and output the power prediction value corresponding to the candidate frequency set through a power estimation model. Based on the power prediction value, the energy-saving potential per unit after adjusting from the current frequency to the candidate frequency is calculated through the energy-saving potential algorithm. The candidate frequencies corresponding to the unit energy-saving potential greater than the threshold are input into the feasible frequency set. The energy-saving potential algorithm includes the estimated value of the average power in the most recent period at the current frequency, the power prediction value of the candidate frequency, and a trend weighting term. The trend weighting term is used to automatically adjust the sensitivity of the energy-saving estimation according to different trend states.

[0024] The target frequency generation module establishes a comprehensive scoring function based on the feasible frequency set and its corresponding unit energy-saving potential, according to the dual constraints of energy-saving efficiency and operational safety. The module scores the feasible frequency set through the comprehensive scoring function to generate a scoring sequence and selects the optimal target frequency from multiple feasible frequencies.

[0025] The frequency correction model is used to input the optimal target frequency and the current operating frequency into the feedback correction function, output the execution frequency of the frequency converter through the feedback correction function, and convert the execution frequency into the corresponding control command. The feedback correction function includes the weighted result of the optimal target frequency and the current operating frequency and the frequency adjustment execution weight factor.

[0026] The indicator update module is used to update energy-saving assessment indicators based on execution frequency, providing dynamic feedback data for trend identification and energy-saving potential prediction in the next cycle.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] This application first uses a multivariate time window to form a load state sequence and performs dynamic trend identification, elevating the current situation to an insight into recent trends, providing a structured entry point for subsequent calculations. Second, under the joint constraints of trend results and the current frequency, the energy-saving potential of candidate frequencies is estimated, and a trend-related response factor is set to ensure differentiated sensitivity of estimates under different trends, avoiding blind frequency reduction in rising scenarios or hesitation in falling scenarios. Third, when selecting the target frequency, energy-saving benefits, frequency change amplitude, and load inertia characterized by the power change rate are considered simultaneously to construct a comprehensive score and superimpose a single-cycle change range. By limiting the scope of the decision-making process, "energy saving" and "safe implementation" are combined into a comparable decision quantity. Then, at the execution level, a score-driven frequency smoothing mechanism is employed, dynamically interpolating the current frequency with the target frequency using weighted factors. This allows the adjustment path to adaptively converge with energy-saving value and operational stability, reducing impact and improving executability. Finally, after frequency adjustment, power is collected to form a weighted smoothing quantity and dynamic energy-saving index. The actual results of this cycle are recorded in a directly usable quantitative form and used for state construction and estimation correction in subsequent cycles, thereby achieving sequential connection and long-term stable optimization of identification, evaluation, decision-making, and execution. The variables in each of these stages are uniformly named, clearly sourced, and the processes are reusable, adaptable to common scenarios such as fans, pumps, and compressors, and possess engineering feasibility for deployment on controllers or edge devices without modifying existing hardware. Attached Figure Description

[0029] Figure 1 This is a flowchart of a variable frequency speed regulation energy-saving control method based on dynamic load identification in a specific embodiment of the present invention;

[0030] Figure 2 This is a block diagram of a variable frequency speed regulation energy-saving control system based on dynamic load identification in a specific embodiment of the present invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] refer to Figure 1 As shown, this application proposes a variable frequency speed regulation energy-saving control method based on dynamic load identification, including:

[0033] Step 1: Construct a multidimensional time series from the collected data within the current system cycle. Input the multidimensional time series into a combined structure of convolutional layers and attention weighting modules. Output trend labels through this combined structure, specifically including:

[0034] This step aims to construct a multi-dimensional time series structure from data collected during the current system operating cycle, and based on this series, identify the dynamic trend of the current load, providing key input information for subsequent energy-saving control strategies. Since this step is the starting point of the control process, all required data comes from real-time acquisition from the system's field equipment. Data sources include: the current output frequency of the frequency converter. Motor current and real-time active power Among them, frequency Current is obtained through the inverter's communication register. The power is collected by a current transformer installed on the motor power supply side, conditioned, and then input to the controller. The sampling period is fixed at 0.1 seconds, which can be obtained through the internal parameters of the frequency converter or an independent power monitoring module to ensure sufficient dynamic response capability.

[0035] The collected data is structured in time series format into a sliding window structure, as follows:

[0036] ;

[0037] in, This indicates a length of [length], with the current time as the endpoint. Multivariable time-series state sequences, including Each time slice contains three observations. The sequence length is... The value is set according to the load response characteristics, and is usually 20, which represents the operating status in the last 2 seconds. Indicates the previous number The frequency of each cycle, This represents the current in the corresponding period. This represents the power of the cycle. All variables are quantitatively collected values, possessing stable physical meaning and engineering reproducibility.

[0038] To extract variation features and determine trends from the sequence, this step designs a combined structure comprising two convolutional layers and an attention-weighted module. The first part is a convolutional extraction unit, which extracts short-term variation patterns from the input sequence. This module contains two one-dimensional convolutional layers: the first layer has 8 kernels, and the second layer has 16 kernels. The kernel width of each layer is set to 3, and the stride is 1. The convolutional output is used to represent the local rate of change and the degree of fluctuation in the state sequence. The convolutional output is then input to the attention-weighted module. This module weights the state representations of different time slices to highlight key time segments in the load state change process. Attention weighting calculates the correlation coefficient between states to form a set of weights; a higher weight indicates a greater influence of that time point on the current trend. All time slices are combined into a single weighted state vector for trend determination.

[0039] The weighted state vector is input into a mapping layer, which outputs three trend classification scores, corresponding to "load will decrease," "load will remain stable," and "load will increase," respectively. The score with the highest value is then used as the identification label for that period. The above calculation process can be expressed in the following form:

[0040] ;

[0041] in, The load trend classification result for the current period, with a value range of [value range missing]. Where 0 indicates a decrease in load, 1 indicates a stable load, and 2 indicates an increase in load; In order to be with the first The score corresponding to each trend category is calculated by the linear mapping relationship between the weighted state vector and each category. This model structure is trained using historical operating data before system deployment. After the parameters are solidified, it is embedded into the industrial edge controller for operation. The inference time is controlled within 100 milliseconds, making it suitable for industrial real-time control scenarios.

[0042] For example, in a certain wind turbine system, if the power gradually increases and the current increases while the frequency remains constant in the collected operating data over the past 2 seconds, the model will output a label. This indicates that "the load will increase," serving as an important basis for subsequent judgments on whether frequency control should be increased.

[0043] Step 2: Establish a candidate frequency set and output the power prediction values ​​corresponding to the candidate frequency set through the power estimation model. Based on the power prediction values, calculate the unit energy-saving potential after adjusting from the current frequency to the candidate frequency using the energy-saving potential algorithm. Input the candidate frequencies corresponding to the unit energy-saving potential greater than the threshold into the feasible frequency set. The energy-saving potential algorithm includes the estimated value of the average power in the most recent period at the current frequency, the power prediction value of the candidate frequency, and a trend weighting term. The trend weighting term is used to automatically adjust the sensitivity of the energy-saving estimation according to different trend states, specifically including:

[0044] The core objective of this step is to leverage the load trend tags identified in the previous step. and historical state sequence Combined with the current operating frequency This step assesses whether adjusting the operating frequency under current conditions would result in significant energy savings, thus providing a reliable and actionable basis for selecting the next frequency regulation strategy. This step holds an irreplaceable position in the entire invention system because it solves the problem of "only identifying, not judging" in traditional variable frequency control. That is, even if past energy-saving control systems knew the load was changing, they could not determine whether frequency adjustment was worthwhile. By coupling the trend identification results with the energy-saving benefit estimation mechanism, this step introduces a future-oriented, goal-driven frequency regulation value judgment mechanism, which is the central link in realizing the core solution of the invention: "variable frequency speed regulation energy-saving control based on dynamic load identification."

[0045] Regarding input, It is the trend label output in the previous step, with a value range of... These represent "load decrease", "load stabilize", and "load increase", respectively. It is a sequence of sliding states constructed in the previous cycle, consisting of the current and historical states. The operating frequency, current, and power composition within each cycle; These are the actual operating frequency values ​​set within the current control cycle, which are set by the control system in the previous cycle and stored in the control register. These three variables are all the input variables required for energy-saving judgment in this step, and they have the characteristics of being clearly sourced, collectable, and having a unique meaning.

[0046] The first stage of this step is to establish a candidate frequency set. This set represents the variable frequency control targets that may be executed within the current cycle, typically including several upper and lower limit frequencies around the current frequency. For example, suppose the current frequency is... Hz, can be constructed The frequency interval (Hz) should be set in conjunction with the specific equipment's response inertia, frequency resolution, and allowable control precision.

[0047] Then, each candidate frequency Substitute the values ​​into a power estimation model trained based on historical operating data, and output the corresponding predicted power value. The power estimation model consists of a linear regression structure superimposed with trend factor weighting terms, and has a two-layer structure: the first layer input is the current state features (such as the average power of the last 5 periods). Average current and current frequency The second layer is related to trend tags. The corresponding trend-weighted adjustment module trains the model offline using historical data before system deployment, and then deploys it on the edge control terminal for real-time inference after the parameters are fixed.

[0048] The innovation of this step lies in the introduction of a trend-weighted term. This is used to automatically adjust the sensitivity of energy-saving estimates based on different trend states. For example, when the load is on an "upward trend" ( Therefore, it is necessary to prioritize avoiding underfrequency issues that could prevent the system from meeting demand. Thus, the estimation model introduces a penalty term to reduce the energy savings corresponding to frequency reduction. When the load is in a "decreasing trend" ( If the frequency reduction strategy is to be encouraged, then the energy-saving response weight of the frequency reduction solution should be increased.

[0049] The specific energy-saving potential algorithm is as follows:

[0050] ;

[0051] in, This indicates that the frequency will be changed from... Adjusted to The potential for energy saving per unit after the process; Based on the most recent frequency at the current frequency Estimated average power per cycle; In candidate frequency The predicted power value is output by the linear regression model and the trend correction module. The trend response factor is defined as follows:

[0052] ;

[0053] in, This is the adjustment coefficient (usually set between 0.1 and 0.3). This is the trend label for the current period. When hour, This indicates that no additional energy-saving value will be increased; while when hour, This means amplifying the energy-saving potential corresponding to the current frequency reduction, encouraging the controller to try frequency reduction strategies. The introduction of this trend-aware factor effectively improves the estimation model's ability to adapt to the dynamic characteristics of actual loads, making frequency adjustment decisions more aligned with energy-saving goals.

[0054] After the energy-saving potential is estimated, for each Perform with energy saving threshold By comparing the frequencies, a feasible frequency set is constructed:

[0055] ;

[0056] in, This is the minimum acceptable energy saving value set by system engineers based on factors such as equipment rated energy consumption and frequency regulation costs; it is typically 1% to 3% of the equipment's full-load power. This represents the set of frequency modulation frequencies that are "worth executing" after energy-saving estimation, trend response weighting, and threshold filtering, based on the current trend assessment.

[0057] In practical deployments, the energy-saving potential assessment module is deployed in the controller or edge computing node, with inference time controlled within 100 milliseconds to ensure the overall system response timeliness. All parameters required by the model are trained before deployment and injected into the controller system via a standard configuration file. The aforementioned calculation process is highly reproducible and structurally interpretable, and is directly coupled with the proposed "load-based dynamic identification" system control concept.

[0058] The output variables for this step include: First, the frequency regulation energy-saving potential sequence. First, it serves as an evaluation metric for the next step of frequency modulation strategy selection; second, it forms a candidate frequency set. This indicates a frequency modulation scheme worth considering in the current cycle.

[0059] Step 3: Based on the set of feasible frequencies and their corresponding unit energy-saving potential, a comprehensive scoring function is established according to the dual constraints of energy efficiency and operational safety. The feasible frequency set is scored using this comprehensive scoring function to generate a scoring sequence, and the optimal target frequency is selected from multiple feasible frequencies. Specifically, this includes:

[0060] The core task of this step is to analyze the obtained candidate frequency set. and its corresponding energy-saving potential valuation Based on this, and taking into account both energy efficiency and operational safety constraints, the optimal target frequency is selected from multiple candidate frequencies. This step serves as the "strategy convergence" function in the entire invention control chain: by comprehensively evaluating the benefits and costs of each frequency modulation strategy, it ensures that the frequency modulation operation ultimately executed by the system is not only energy-efficient but also complies with equipment protection requirements, system response patterns, and operating condition limitations. It builds upon the "energy-saving potential space" established in step two, transforming trend-aware energy-saving judgments into practically executable frequency modulation targets, representing a crucial leap "from analysis to control."

[0061] In industrial scenarios, such as large fan and water pump load systems, rapid or large frequency changes can easily lead to mechanical shock, water hammer effect, or excessively rapid motor temperature rise, affecting system lifespan and operational stability. Therefore, this step innovatively introduces two safety awareness items when constructing the frequency scoring index: one is a frequency modulation amplitude penalty item, and the other is a load inertia response item, used to finely characterize the "physical cost" of frequency change behavior.

[0062] Let the comprehensive scoring index be The comprehensive scoring function is defined as follows:

[0063] ;

[0064] in, Candidate frequency The overall optimal score is the weighted difference between energy-saving benefits and frequency regulation costs; It is the energy-saving potential value estimated in step two; the first penalty factor The impact of frequency adjustment amplitude on the score is controlled; the numerical setting range is typically 0.1 to 0.5, depending on the actual allowable frequency modulation rate of the equipment; the second item is the penalty factor. This is used to control the risks caused by load response inertia. It represents the rate of change of the current average power, and its squared form is used to amplify the risk of frequency fluctuations under sudden load changes.

[0065] in, It is the average power within the sliding window, derived from... Middle and near The power data at each time point is calculated using a simple arithmetic mean. Its rate of change... It is obtained by dividing the difference between the current and previous period's mean by the control period time interval.

[0066] For example, for a system with a sampling frequency of 10Hz and a control period of 0.1 seconds, if the average value of the previous period was 2.0kW and the average value of the current period is 2.4kW, then the rate of change is... Its square is 16.0 kW² / s². This item can significantly reduce the frequency score of some high-energy-saving but unstable frequencies under drastic load fluctuations, ensuring frequency regulation safety.

[0067] In addition, to ensure the physical feasibility of frequency modulation, it is also necessary to set limits on the frequency variation range:

[0068] ;

[0069] in, The maximum allowable single-cycle frequency variation for system operation is typically set to 2Hz to 5Hz. Frequencies exceeding this limit will be... Those removed from the list will not be included in the scoring and ranking.

[0070] Ultimately, all candidate frequencies that meet the physical constraints of frequency modulation are... According to its The values ​​are sorted, and the one with the highest score is selected as the target frequency for the current cycle. :

[0071] ;

[0072] This formula ensures that the ultimately selected target frequency not only possesses theoretical energy-saving potential but also has the physical justification for safe implementation. Compared to the traditional control strategy's method of simply ranking by energy saving potential, this step introduces a multi-dimensional constraint optimization mechanism, incorporating energy-saving potential, safety costs, and load inertia into the frequency regulation strategy scoring, forming a structured control decision criterion.

[0073] For example, in an urban water supply pump system, if the current frequency is 48Hz, the candidate frequency set is: Hz, energy saving potential respectively kW, power change rate kW / s, assuming , The scores for the three are as follows:

[0074] ;

[0075] ;

[0076] ;

[0077] although It has the greatest corresponding energy-saving potential, but its load fluctuation cost is also the highest. Although the energy savings were relatively small, the overall score was high, therefore this cycle selected... Hz is the frequency modulation target.

[0078] Step 4: Input the optimal target frequency and the current operating frequency into the feedback correction function. The feedback correction function outputs the inverter's execution frequency and converts it into corresponding control commands. The feedback correction function includes a weighted result of the optimal target frequency and the current operating frequency, as well as a frequency modulation execution weighting factor, specifically including:

[0079] This step is the most critical execution stage in the entire variable frequency speed control energy-saving scheme. Its task is to convert the optimal target frequency output from the previous step into the desired frequency. In practical applications, this method is used in inverter control interfaces to dynamically adjust motor speed based on prediction, judgment, and selection results. Unlike the traditional "direct frequency setting" method, this step is based on multi-dimensional information obtained in previous steps (including the target frequency). Frequency rating Historical operating frequency A multi-factor dynamic frequency regulation execution mechanism was constructed, which includes power state changes. This mechanism not only realizes the physical control of frequency setting, but also introduces dynamic response adjustment weights, system risk balance functions, and trend prediction feedback smoothing factors to make the control behavior more driven by energy-saving goals and safer for load response. It effectively solves the problem of "mismatch between energy-saving behavior and physical risk" and is the core interface for the implementation of the entire control method.

[0080] The input includes: the optimal target frequency output from step three. Current operating frequency and the comprehensive score corresponding to the target frequency. .in, The frequency setting value is read directly from the system status cache and is sent to the inverter in the previous control cycle. The value, calculated from the optimized model in the previous step, reflects the overall value of the current frequency regulation operation (a higher value indicates greater energy savings and better system adaptability). Additionally, historical state sequences need to be incorporated. Nearly extracted Average active power over one cycle and its rate of change This serves as the input basis for evaluating the system's response inertia and stability.

[0081] Traditional frequency conversion execution strategies typically directly... While writing settings into the inverter's configuration channel is feasible, in industrial scenarios with drastic load responses or sensitive equipment structures (such as large centrifugal pumps and heavy-duty fans), this strategy can easily lead to current fluctuations, excessive impacts, or trigger system interlocking protection, causing the energy-saving strategy to actually cause system instability. This step proposes a dynamic interpolation frequency regulation mechanism based on a multi-dimensional feedback correction function, constructing a frequency setting strategy that is "physically controllable, energy-saving behavior-oriented, and response behavior smooth."

[0082] The core calculation logic is as follows:

[0083] ;

[0084] in, The final execution frequency sent to the inverter is and The dynamic weighted result between them; It is the frequency modulation execution weighting factor, reflecting the "aggressiveness" of frequency modulation allowed by the current system, and its value range is... . A larger value indicates that the frequency is rapidly approaching the target, while a smaller value indicates that the frequency is adjusted slowly to ensure system stability.

[0085] The calculation method fully integrates energy-saving score and load trend label. Current power change rate Innovatively, a weighted suppression function is constructed by introducing an inertia suppression term and a trend adaptation term, taking into account various factors. ,as follows:

[0086] ;

[0087] in: The overall score for the current target frequency. The maximum value in the scoring sequence for this period represents the relative score. To prevent extremely small positive numbers from being divided by zero; It represents the rate of change of the current load power, used to measure the inertia of load changes; For trend tags, the value is... , respectively representing "decline", "stability", and "rise", among which This indicates the "degree of encouragement for frequency modulation"; the value is highest when the trend is downward, which helps to amplify the weight of frequency modulation. , This is an empirical adjustment coefficient that controls the intensity of inertia penalty and trend guidance, and is generally taken in the range of 0.1 to 0.5. This is the weighting correction coefficient; the larger its value, the more the current system encourages rapid frequency adjustment.

[0088] The innovation of this formula lies in the fact that it no longer uses "whether it is energy-saving" as the sole condition for frequency regulation. Instead, it introduces an adaptive frequency regulation weight control mechanism with two factors, physical and trend, to map the system state to the frequency regulation speed, thus balancing the conflict between regulation response and stability.

[0089] For example, when the system has a high energy efficiency score, but the load changes too drastically ( When (too large), It will decay rapidly, suppressing the frequency modulation speed and preventing system instability; however, when the load trend is clear, the fluctuations are small, and the energy-saving score is high, The trend term enhances the system's ability to quickly adjust its frequency, thereby improving the responsiveness of energy-saving benefits.

[0090] Final Execution Frequency This will be translated into corresponding control commands, the specific implementation of which depends on the system configuration:

[0091] If the system adopts analog control (such as 4) 20mA或0 (10V output), then through proportional conversion to Mapped to analog voltage or current;

[0092] If digital communication methods (such as Modbus RTU, CANopen, Profibus, etc.) are used, then Convert the value to the setting register and write it to the inverter control address (such as Schneider inverters). register);

[0093] All control signals are executed by the system's main PLC or industrial edge controller via a communication module. Before execution, the frequency is automatically checked to ensure it does not exceed the physical upper or lower limits. .

[0094] Step 5: Collect the actual operating power corresponding to the execution frequency and update the energy-saving assessment indicators to provide dynamic feedback data for trend identification and energy-saving potential prediction in the next cycle. Specifically, this includes:

[0095] The task of this step is to collect the actual operating power and update the energy-saving assessment indicators after the variable frequency speed control operation is completed (output of step four). This provides dynamic feedback data for trend identification and energy-saving potential prediction in the next cycle, and is a key link in achieving closed-loop optimization of the entire system. Actual power data The power is collected by a three-phase power sensor installed at the motor output. The sensor can directly output a digital signal, which is input to the controller via RS485 or CAN bus. Alternatively, it can output an analog signal, which is converted by an A / D module before entering the control system. To ensure data comparability, this step uses the "weighted real-time power" method, which compares the instantaneous power value with the frequency modulation weighting factor. Combined to form a smoother and trend-aware power assessment metric. :

[0096] ;

[0097] in, The power value after weighted smoothing for the current period is a representation of the true power. Smooth power compared to the previous cycle The weighted result. The larger the value, the more aggressive the frequency modulation is, and the higher the weighting percentage of the real-time power value in the current cycle. The smaller the value, the closer the smoothed power is to historical values, avoiding misleading the model due to drastic power fluctuations. This formula allows the system to incorporate the aggressiveness of frequency adjustments into power assessment, forming a more robust dynamic power benchmark.

[0098] To quantify energy-saving effects, this step also introduces a "dynamic energy-saving assessment index". This is used to measure the improvement in power after the current frequency modulation is performed relative to the reference power. Reference power The dynamic energy-saving assessment index can be established based on the average power consumption under historical load trends and frequencies. It is initialized with historical data during the initial system deployment and gradually updated during operation. The dynamic energy-saving assessment index is defined as follows:

[0099] ;

[0100] in, This indicates the proportion of energy savings achieved by the current frequency modulation strategy; a larger value indicates more significant energy savings. The reference power is derived from historical data of similar operating conditions or long-term statistical values ​​of the system. To prevent extremely small positive numbers with a denominator of zero, this indicator is calculated and stored in each cycle, and used as a dynamic correction quantity for the trend identification model and energy-saving potential prediction model in the next cycle. This forms a "self-learning, self-updating" energy-saving effect database, which significantly improves the prediction accuracy and reliability of the entire system's energy-saving strategy.

[0101] For example, in a large water pump system, if the smoothed power of the previous cycle... kW, frequency modulation weight for this cycle Real-time power measurement kW, then kW; if the reference power kW, then This indicates that the current frequency regulation strategy saves approximately 5.8% of energy.

[0102] refer to Figure 2 As shown, in a second aspect of this application, a variable frequency speed control energy-saving control system based on dynamic load identification is also proposed, comprising:

[0103] The trend recognition module is used to construct a multidimensional time series from the data collected during the current operating cycle of the system. The multidimensional time series is input into a combined structure of convolutional layer and attention weighting module, and trend labels are output through the combined structure.

[0104] The energy-saving potential assessment module is used to establish a candidate frequency set and output the power prediction value corresponding to the candidate frequency set through a power estimation model. Based on the power prediction value, the energy-saving potential per unit after adjusting from the current frequency to the candidate frequency is calculated through the energy-saving potential algorithm. The candidate frequencies corresponding to the unit energy-saving potential greater than the threshold are input into the feasible frequency set. The energy-saving potential algorithm includes the estimated value of the average power in the most recent period at the current frequency, the power prediction value of the candidate frequency, and a trend weighting term. The trend weighting term is used to automatically adjust the sensitivity of the energy-saving estimation according to different trend states.

[0105] The target frequency generation module establishes a comprehensive scoring function based on the feasible frequency set and its corresponding unit energy-saving potential, according to the dual constraints of energy-saving efficiency and operational safety. The module scores the feasible frequency set through the comprehensive scoring function to generate a scoring sequence and selects the optimal target frequency from multiple feasible frequencies.

[0106] The frequency correction model is used to input the optimal target frequency and the current operating frequency into the feedback correction function, output the execution frequency of the frequency converter through the feedback correction function, and convert the execution frequency into the corresponding control command. The feedback correction function includes the weighted result of the optimal target frequency and the current operating frequency and the frequency adjustment execution weight factor.

[0107] The indicator update module is used to update energy-saving assessment indicators based on execution frequency, providing dynamic feedback data for trend identification and energy-saving potential prediction in the next cycle.

[0108] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A variable frequency speed regulation energy-saving control method based on dynamic load identification, characterized in that, include: A multidimensional time series is constructed from the collected data during the current system cycle. The multidimensional time series is then input into a combined structure of convolutional layers and attention weighting modules, and trend labels are output through the combined structure. A candidate frequency set is established, and the power prediction value corresponding to the candidate frequency set is output through the power estimation model. Based on the power prediction value, the unit energy saving potential after adjusting from the current frequency to the candidate frequency is calculated through the energy saving potential algorithm. The candidate frequencies corresponding to the unit energy saving potential greater than the threshold are input into the feasible frequency set. The energy saving potential algorithm includes the estimated value of the average power in the most recent period at the current frequency, the power prediction value of the candidate frequency, and a trend weighting term. The trend weighting term is used to automatically adjust the sensitivity of the energy saving estimation according to different trend states. Based on the set of feasible frequencies and their corresponding unit energy-saving potential, a comprehensive scoring function is established according to the dual constraints of energy-saving efficiency and operational safety. The set of feasible frequencies is scored by the comprehensive scoring function to generate a scoring sequence, and the optimal target frequency is selected from multiple feasible frequencies. The optimal target frequency and the current operating frequency are input into the feedback correction function, the execution frequency of the frequency converter is output through the feedback correction function, and the execution frequency is converted into the corresponding control command. The feedback correction function includes the weighted result of the optimal target frequency and the current operating frequency and the frequency adjustment execution weight factor. The system collects the actual operating power corresponding to the execution frequency and updates the energy-saving assessment indicators, providing dynamic feedback data for trend identification and energy-saving potential prediction in the next cycle.

2. The variable frequency speed regulation energy-saving control method based on dynamic load identification according to claim 1, characterized in that, The process of converting the execution frequency into corresponding control commands specifically includes: If the system uses analog control, the execution frequency is mapped to analog voltage or current through proportional conversion; If digital communication is used, the execution frequency is converted into a setting register value and written to the inverter control address.

3. The variable frequency speed regulation energy-saving control method based on dynamic load identification according to claim 1, characterized in that, The comprehensive scoring function includes unit energy saving potential, a first penalty factor, a second penalty factor, and the rate of change of the current average power. The first penalty factor is used to control the impact of frequency adjustment on the score, and the second penalty factor is used to control the risk caused by load response inertia.

4. The variable frequency speed regulation energy-saving control method based on dynamic load identification according to claim 1, characterized in that, The frequency modulation execution weighting factor is used to map the system state to the frequency modulation speed, balancing the conflict between regulation response and stability. The frequency modulation execution weighting factor includes the comprehensive score corresponding to the optimal target frequency, the smallest positive number to prevent division by zero, the power change rate of the current load, the trend label, the empirical regulation coefficient, and the weight correction coefficient.

5. The variable frequency speed regulation energy-saving control method based on dynamic load identification according to claim 1, characterized in that, The energy-saving assessment index represents the proportion of energy-saving effect brought about by the current frequency regulation strategy. The energy-saving assessment index is generated by an assessment function, which includes a reference power, a very small positive number to prevent the denominator from being zero, and a power assessment index.

6. The variable frequency speed regulation energy-saving control method based on dynamic load identification according to claim 5, characterized in that, The power evaluation index is generated by combining the instantaneous power value with the frequency modulation execution weighting factor. The instantaneous power value is collected by a three-phase power sensor installed at the motor output.

7. The variable frequency speed regulation energy-saving control method based on dynamic load identification according to claim 1, characterized in that, The power estimation model consists of a linear regression structure superimposed with trend factor weighting terms.

8. The variable frequency speed regulation energy-saving control method based on dynamic load identification according to claim 1, characterized in that, The trend labels include load decreasing, load stabilizing, and load increasing.

9. The variable frequency speed regulation energy-saving control method based on dynamic load identification according to claim 1, characterized in that, The collected data includes the inverter's current operating frequency, motor current, and real-time active power. The current operating frequency is obtained through the inverter's communication register. The motor current is collected by a current transformer installed on the motor's power supply side, and after signal conditioning, it is input to the controller. The real-time active power can be obtained through the inverter's internal parameters or an independent power monitoring module.

10. A variable frequency speed regulation energy-saving control system based on dynamic load identification, characterized in that, include: The trend recognition module is used to construct a multidimensional time series from the data collected during the current operating cycle of the system. The multidimensional time series is input into a combined structure of convolutional layer and attention weighting module, and trend labels are output through the combined structure. The energy-saving potential assessment module is used to establish a candidate frequency set and output the power prediction value corresponding to the candidate frequency set through a power estimation model. Based on the power prediction value, the energy-saving potential per unit after adjusting from the current frequency to the candidate frequency is calculated through the energy-saving potential algorithm. The candidate frequencies corresponding to the unit energy-saving potential greater than the threshold are input into the feasible frequency set. The energy-saving potential algorithm includes the estimated value of the average power in the most recent period at the current frequency, the power prediction value of the candidate frequency, and a trend weighting term. The trend weighting term is used to automatically adjust the sensitivity of the energy-saving estimation according to different trend states. The target frequency generation module establishes a comprehensive scoring function based on the feasible frequency set and its corresponding unit energy-saving potential, according to the dual constraints of energy-saving efficiency and operational safety. The module scores the feasible frequency set through the comprehensive scoring function to generate a scoring sequence and selects the optimal target frequency from multiple feasible frequencies. The frequency correction model is used to input the optimal target frequency and the current operating frequency into the feedback correction function, output the execution frequency of the frequency converter through the feedback correction function, and convert the execution frequency into the corresponding control command. The feedback correction function includes the weighted result of the optimal target frequency and the current operating frequency and the frequency adjustment execution weight factor. The indicator update module is used to update energy-saving assessment indicators based on execution frequency, providing dynamic feedback data for trend identification and energy-saving potential prediction in the next cycle.

Images