An electrical equipment intelligent efficiency evaluation and management system for a photovoltaic power station

Abstract

The application discloses a photovoltaic power station-oriented electrical equipment intelligent efficiency evaluation and management system, and belongs to the technical field of data processing.The system comprises a data acquisition module, which is used for acquiring environmental data, electrical data and running state data of a target electrical equipment in real time; a state evaluation module, which is used for determining a function steady state deviation of the target electrical equipment based on the data acquired by the data acquisition module and in combination with a preset function steady state interval; a decision optimization module, which is used for generating an optimal control instruction according to the function steady state deviation and a preset reward function; and a control execution module, which is used for executing the optimal control instruction to regulate and control the running state of the target electrical equipment.The application establishes a quantifiable equipment health state evaluation system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, in particular to an intelligent efficiency evaluation and management system for electrical equipment of a photovoltaic power station. BACKGROUND

[0002] In the conventional operation of a photovoltaic power station, the core goal of the management system is generally focused on maximizing instantaneous power generation efficiency and ensuring the reliability of basic operation. Technical solutions mostly use algorithms such as maximum power point tracking, which are committed to extracting maximum power under various environmental conditions. However, this aggressive control strategy, which pursues short-term power generation as the only goal, inevitably leads to the inverter and other key power electronic devices bearing severe electrical and thermal stress cycles. Long-term stress accumulation can accelerate the aging process of internal power semiconductors, capacitors, and other components of the equipment, shortening their effective service life and increasing long-term maintenance costs. Existing technical solutions often perform passive evaluation and maintenance only after the performance of the equipment has significantly declined, lacking a mechanism to predict and actively manage the health status of the equipment. Therefore, how to maintain the equipment in the healthiest working state through active intervention while ensuring reasonable power generation income to maximize the delay of its performance decline is a technical problem that needs to be solved in the field. SUMMARY

[0003] The purpose of the present application is to provide an intelligent efficiency evaluation and management system for electrical equipment of a photovoltaic power station to solve the problems raised in the background.

[0004] The technical solution of the present application is as follows:

[0005] A data acquisition module for real-time acquisition of environmental data, electrical data, and operating state data of the target electrical equipment;

[0006] A state evaluation module for determining the functional steady state deviation of the target electrical equipment based on the data acquired by the data acquisition module and in combination with the preset functional steady state interval;

[0007] A decision optimization module for generating optimal control instructions according to the functional steady state deviation and the preset reward function;

[0008] A control execution module for executing the optimal control instructions to regulate the operating state of the target electrical equipment.

[0009] Preferably, the state evaluation module determines the functional steady state deviation, comprising:

[0010] extract a stress indicator current value from data acquired by a data acquisition module; call a preset functional steady state interval, the functional steady state interval being jointly defined by a preset stress indicator optimal central value and a preset stress indicator safety range; call a preset weight coefficient; based on the stress indicator current value, the functional steady state interval and the weight coefficient, calculate and generate a functional steady state deviation degree.

[0011] Preferably, the determination of the weight coefficient comprises:

[0012] In the preset digital twin model, each stress indicator is subjected to systematic disturbance; through accelerated simulation, the negative influence degree of each stress indicator on the predicted residual service life of the equipment is quantified; the weight coefficient corresponding to each stress indicator is determined according to the normalized processing of the negative influence degree.

[0013] Preferably, the decision optimization module generates an optimal control instruction, comprising:

[0014] The functional steady state deviation degree is taken as a negative penalty term in a preset reward function; based on the preset reward function, a deep reinforcement learning task is used for optimization to generate an optimal control instruction for intervention on the target electrical equipment.

[0015] Preferably, the preset reward function comprises:

[0016] An instantaneous power generation efficiency term representing power generation income;

[0017] A functional steady state deviation degree representing the health status of the equipment;

[0018] A power grid violation penalty term ensuring power grid friendliness.

[0019] Preferably, the determination of the instantaneous power generation efficiency term comprises:

[0020] An actual output power provided by the data acquisition module is acquired; a preset digital twin model is called to calculate a theoretical maximum power under current environmental conditions; the instantaneous power generation efficiency term is determined by calculating the ratio of the actual output power to the theoretical maximum power.

[0021] Preferably, the determination of the power grid violation penalty term comprises:

[0022] The output indicator of the target electrical equipment is compared with a preset power grid specification;

[0023] When the output indicator is out of limit, the power grid violation penalty term is assigned a preset negative value;

[0024] When the output indicator is not out of limit, the power grid violation penalty term is assigned a value of zero.

[0025] Preferably, it further comprises:

[0026] The performance evaluation module is configured to determine a predicted lifetime under a steady-state control strategy constituted by the optimal control instruction based on the preset digital twin model, and determine a predicted lifetime under a standard control strategy, and obtain a degradation suppression factor representing long-term benefits of the steady-state control strategy by calculating a ratio of the two.

[0027] Preferably, the performance evaluation module is further configured to:

[0028] determine an actual output power guided by the optimal control instruction under the method;

[0029] determine a theoretical maximum power under a current environmental condition;

[0030] calculate an integral of a difference between the actual output power and the theoretical maximum power within a preset time;

[0031] calculate an integral of the theoretical maximum power within the preset time;

[0032] determine a regulation efficiency loss rate representing short-term costs of the steady-state control strategy by dividing the integral of the difference by the integral of the theoretical maximum power

[0033] The present application provides an intelligent performance evaluation and management system for electrical equipment of a photovoltaic power station by improvement, and has the following improvements and advantages compared with the prior art.

[0034] 1. The present scheme establishes a quantifiable equipment health state evaluation system; through the state evaluation module, the concept of function steady-state deviation degree is introduced, and stronger punishment is applied to greater health deviation; the function steady-state deviation degree calculated by combining the stress index current value provided by the data acquisition module and the preset function steady-state interval makes the abstract equipment health first quantified as an accurate, dimensionless scalar, providing a solid mathematical foundation for subsequent closed-loop control;

[0035] 2. The control priority of the present scheme has a clear physical basis, ensuring that the distribution of weight coefficients is directly related to the real physical degradation mechanism, so that the decision optimization module can preferentially suppress the most critical damage factors, greatly improving the pertinence and efficiency of health management, which is unmatched by the experience-based and indiscriminate control methods in the prior art;

[0036] 3. The present scheme realizes dynamic optimal balance of power generation income and equipment health, ensuring that all optimization behaviors are within the boundary of power grid safety specifications; compared with the single and static control target of the prior art, the agent of the present scheme can autonomously learn and execute a complex and dynamic optimal control instruction to maximize long-term cumulative rewards, thereby achieving Pareto optimality of power generation income and equipment health in practice. BRIEF DESCRIPTION OF DRAWINGS

[0037] The application will be further explained in connection with the accompanying drawings and embodiments:

[0038] Figure 1 is a flow chart of an electrical equipment intelligent efficiency evaluation and management system for a photovoltaic power station of the application;

[0039] Figure 2 is a function steady state deviation determination step diagram in embodiment 2. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical scheme and advantages of the application more clear, the application will be further described in detail below in combination with specific embodiments.

[0041] Embodiment 1:

[0042] Please refer to Figure 1 The application provides an electrical equipment intelligent efficiency evaluation and management system for a photovoltaic power station, comprising:

[0043] A data acquisition module is configured to acquire environmental data, electrical data and running state data of a target electrical equipment in real time.

[0044] A state evaluation module is configured to determine a function steady state deviation of the target electrical equipment based on the data acquired by the data acquisition module and in combination with a preset function steady state interval.

[0045] A decision optimization module is configured to generate an optimal control instruction according to the function steady state deviation and a preset reward function.

[0046] A control execution module is configured to execute the optimal control instruction to regulate the running state of the target electrical equipment.

[0047] The application provides an electrical equipment intelligent efficiency evaluation and management system for a photovoltaic power station; the system aims to solve the problem that the traditional control strategy pursues the maximum instantaneous power generation at the cost of sacrificing the long-term health of the equipment through proactive active health management, so as to maximize the delay of the performance degradation of the key electrical equipment while ensuring reasonable power generation income and reducing long-term operation and maintenance cost.

[0048] In the present embodiment, the system is deployed to manage the inverters in the photovoltaic power station; the system is constructed as a closed-loop regulation system, and the composition and working process are described.

[0049] A data acquisition module is configured to acquire environmental data, electrical data, and operating state data of the target electrical device in real time. The environmental data refers to physical quantities reflecting the working conditions of the device, such as the intensity of illumination and the ambient temperature collected by a temperature sensor. The electrical data refers to parameters representing the electrical state inside the device, such as the DC voltage and current of the inverter and the AC output voltage and frequency. The operating state data refers to cumulative information reflecting the historical working conditions of the device, such as the cumulative operating hours or the cumulative power generation. The data acquisition is multi-time scale, including high-frequency data of milliseconds to capture transient stress and low-frequency data of minutes or hours to track long-term trends.

[0050] A state evaluation module is configured to quantitatively evaluate the current health state of the target electrical device based on the data acquired by the data acquisition module and in combination with a preset functional steady state interval. The functional steady state interval refers to the optimal operating range of a set of key electrical and thermodynamic indicators, in which the aging rate of the device caused by stress accumulation is minimized. The interval is not arbitrary, and the internal logic is based on the specifications provided by the device manufacturer, a large amount of laboratory aging acceleration experiment data, and simulation analysis of device aging mechanism models.

[0051] A decision optimization module is configured to generate optimal intervention instructions to guide the device to return to or remain in a healthy state according to the current device health state. The module is configured to generate optimal control instructions based on the functional steady state deviation calculated by the state evaluation module and a preset reward function. The reward function is a mathematical construct for target quantification, which unifies the two conflicting goals of power generation income and device health loss into a single scalar evaluation signal, thereby providing a clear optimization direction for the optimization algorithm.

[0052] A control execution module is configured to convert the decision instructions into actual regulation of the physical device. The module is configured to execute the optimal control instructions generated by the decision optimization module to regulate the operating state of the target electrical device. For example, the module sends instructions to the central processing module of the inverter to fine-tune the perturbation step of the maximum power point tracking algorithm or limit the change slope of the power output.

[0053] Through the cooperative work of the above modules, a complete technical link from data perception, state quantification, intelligent decision-making to closed-loop control is formed.

[0054] By constructing a closed-loop health management system of data acquisition-state evaluation-decision optimization-control execution, the mode transformation from traditional passive response maintenance to proactive active health management is realized; the system can actively intervene and regulate the operation state of the electrical equipment under the premise of ensuring reasonable power generation income, so that the equipment can be kept in the function steady state interval with the lowest aging rate as much as possible, thereby effectively delaying the performance degradation of the equipment, prolonging the effective service life of the equipment, and reducing the total operation and maintenance cost of the entire photovoltaic power station in the life cycle;

[0055] The beneficial effects produced by the embodiments of the present application are rooted in the new technical paradigm constructed by the embodiments, from passive response evaluation to proactive active health management; compared with the prior art which only focuses on the single goal of maximizing instantaneous power generation efficiency, the present scheme makes significant progress in technical ideas, implementation paths and final effects;

[0056] The fundamental limitation of the prior art is that the aggressive control strategy inevitably leads to severe electrical and thermal stress cycles of key electrical equipment such as inverters, accelerating the aging process until performance degradation occurs, and then passive maintenance is performed; the present scheme realizes proactive and active management of the health state of the equipment by introducing closed-loop regulation.

[0057] Embodiment 2

[0058] Please refer to Figure 2 The state evaluation module determines the function steady state deviation degree, comprising:

[0059] The current value of the stress index is extracted from the data obtained by the data acquisition module; a preset function steady state interval is called, which is jointly defined by a preset optimal central value of the stress index and a preset safety range of the stress index; a preset weight coefficient is called; based on the current value of the stress index, the function steady state interval and the weight coefficient, the function steady state deviation degree is calculated and generated;

[0060] The present embodiment limits the process of determining the function steady state deviation degree by the state evaluation module; this process aims to integrate the stress state of multiple dimensions and different physical properties into a unified and quantifiable dimensionless health state measure;

[0061] The process of determining the function steady state deviation degree by the state evaluation module comprises the following steps:

[0062] The current value of the stress index is extracted from the real-time data stream obtained by the data acquisition module; Stress indicator It refers to a physical quantity that directly or indirectly reflects the stress of key components of the equipment, such as power semiconductors IGBT and capacitors, which in the present embodiment can be specifically the junction temperature of IGBT and the ripple current of DC bus capacitor;

[0063] The systematic selection of stress indicators should be based on a failure mode and effects analysis of the target electrical equipment. FMEA should be used to identify the most critical failure modes that have the greatest impact on equipment reliability. The physical stress sources that cause these failure modes, such as temperature, current, and voltage, should then be identified as stress indicators that must be included in the assessment to ensure a comprehensive and targeted assessment.

[0064] The current values ​​of stress indicators are the measured values ​​of these physical quantities at the current moment;

[0065] Call the preset functional steady-state interval; as mentioned above, the functional steady-state interval is defined by the preset optimal center value of the stress index and the preset safety range of the stress index. The optimal center value of the stress index refers to the ideal operating point at which the equipment aging rate is theoretically lowest; the safety range of the stress index refers to the allowable fluctuation range around the center value. Both parameters are pre-set based on the equipment aging mechanism model and experimental data.

[0066] An example of the quantitative determination steps is: for a specific stress indicator, such as IGBT junction temperature, multiple sets of accelerated aging experiments are conducted in the laboratory at different constant temperatures; based on the experimental data, a curve is fitted to show the relationship between the stress and the predicted life of the device; the point on the curve corresponding to the longest life is the optimal center value of the stress The stress range corresponding to when the lifespan drops to 95% of the maximum lifespan is defined as the safety range of the stress ;

[0067] Call the preset weight coefficient; the weight coefficient refers to a set of numerical values, which is used to characterize the relative importance of different types of stress on the long-term health of the equipment;

[0068] Based on the current value of the stress index, the functional steady-state interval and the weight coefficient, the functional steady-state deviation is calculated and generated; in this embodiment, the calculation is completed by a specific mathematical formula; for further explanation, the functional steady-state deviation is introduced The calculation formula is:

[0069]

[0070] in, is the deviation of functional steady state; is the index of the summation term, representing the i-th stress index; n is the total number of stress indices involved in the calculation; For the The weight coefficients of the stresses are derived from the digital twin simulation described in the subsequent embodiments; is the current value of the stress index obtained in real time by the data acquisition module; It is the optimal center value of stress index pre-set based on equipment mechanism and experimental data; for the same pre-set stress index safety range;

[0071] By introducing a comprehensive functional steady-state deviation calculation formula, the complex and multi-dimensional equipment health status is accurately quantified into a single and dimensionless scalar index ; This quantification not only makes the equipment health status intuitive and comparable, but more importantly, provides a clear and explicit optimization target for the subsequent decision optimization module, i.e., minimizing , thereby greatly improving the accuracy and effectiveness of the closed-loop health management system;

[0072] The present scheme establishes a quantifiable equipment health status evaluation system; through the state evaluation module, the concept of functional steady-state deviation is introduced; to concretize this concept, a comprehensive mathematical construction is proposed: , wherein, : steady-state deviation, : weight coefficient, : current value of stress, : optimal value of stress, : stress safety range, : total number of stresses; This formula is not a simple empirical fitting, and its physical meaning lies in that it represents a weighted Euclidean distance between the current equipment state point and the theoretical optimal health center point in the multi-dimensional stress state space; wherein, A dimensionless treatment is performed on different dimensional stress indicators, making them have additivity; The design of the square term makes the deviation grow quadratically, thereby imposing stronger punishment on larger health deviation; The functional steady-state deviation calculated from the current value of the stress indicator provided by the data acquisition module, combined with the pre-set functional steady-state interval, makes the abstract equipment health first quantified into an accurate and dimensionless scalar, providing a solid mathematical foundation for the subsequent closed-loop control.

[0073] Embodiment 3

[0074] The determination of the weight coefficient includes:

[0075] In the pre-set digital twin model, systematic perturbations are performed on each stress indicator; Through accelerated simulation, the negative impact of each stress indicator on the predicted remaining useful life of the equipment is quantified; According to the negative impact degree, the weight coefficient corresponding to each stress indicator is determined by normalization processing;

[0076] The present embodiment further limits the determination method of the weight coefficient ; This method aims to ensure that the allocation of the weight coefficient has a clear physical meaning and engineering basis, rather than being subjective;

[0077] The determination process of the weight coefficient is usually performed offline, and the working principle is to quantify the contribution of various stresses to the life consumption of the equipment through digital twin simulation, including the following steps:

[0078] In a preset digital twin model, each stress index is systematically disturbed; the preset digital twin model is a highly faithful digital counterpart of the physical entity equipment, which integrates the electrical model, the thermodynamic model, and a key mechanism and data-driven multi-scale aging model of the equipment; the aging model can predict the performance degradation trajectory and the remaining useful life of the key components of the equipment according to the input stress profile;

[0079] To enable those skilled in the art to implement, the digital twin model construction method here is as follows:

[0080] Thermodynamic model: a fourth-order Cauer thermal network model is used to describe the transient thermal impedance of the IGBT module, and the parameters of the model are obtained by consulting the manufacturer's data manual or through experimental calibration;

[0081] Aging model: for IGBT power cycle aging, the Norris-Landzberg model combined with temperature cycle amplitude and average junction temperature influence is used to predict its life. For capacitor aging, an Arrhenius equation-based model is used to correlate the relationship between its ripple current, operating temperature and life decay;

[0082] Model calibration: parameterize the above mechanism model, and use optimization algorithms such as particle swarm optimization or genetic algorithm to fit the model simulation output with the real data collected in the laboratory accelerated aging test, and accurately calibrate the model parameters to ensure high fidelity;

[0083] In the digital twin environment, the negative impact of each stress index on the predicted remaining useful life of the equipment is quantified through accelerated simulation; the specific operation is as follows: select the first stress index to produce a unit disturbance while keeping other stress indexes stable at the optimal central value; run the aging model for accelerated simulation to obtain a predicted RUL reduction caused by the stress disturbance; repeat the process for all stress indexes;

[0084] According to the negative impact degree, the weight coefficient corresponding to each stress index is determined ; the RUL reduction caused by each stress index is sorted and normalized to make the stress with the greatest impact on RUL obtain the highest weight coefficient;

[0085] An objective and reproducible weight coefficient determination method is provided; by performing sensitivity analysis in the digital twin model, the weight coefficient directly linked to its actual impact on the predicted service life of the equipment; this ensures that the control system can suppress those most critical damage factors with higher priority, so that the health management strategy is more targeted and the resource allocation is more reasonable, thereby more efficiently delaying the overall aging process of the equipment;

[0086] The control priority of the present scheme has a clear physical basis; the weight coefficient in the above formula its determination process is deeply dependent on a preset digital twin model; by systematically perturbing each stress indicator in the model and simulating its negative impact on the predicted remaining useful life of the equipment, the contribution of different stresses to the aging of the equipment can be quantified; this ensures that the allocation of the weight coefficient is directly related to the real physical degradation mechanism, so that the decision optimization module can preferentially suppress the most critical damage factors, greatly improving the targeting and efficiency of health management, which is unmatched by the existing technology based on experience and indiscriminate control methods.

[0087] Embodiment 4

[0088] The decision optimization module generates optimal control instructions, including:

[0089] The functional steady-state deviation degree is taken as a negative penalty term in the preset reward function; based on the preset reward function, a deep reinforcement learning task is used for optimization to generate optimal control instructions for intervention on the target electrical equipment;

[0090] In the present embodiment, a soft actor-critic algorithm suitable for continuous action space control is specifically used, wherein:

[0091] The network structure of the agent: both the actor network and the critic network use a multilayer perceptron containing three hidden layers, the input layer receives state data, the output layer outputs control instructions, for example, the MPPT perturbation step, and the hidden layer uses ReLU as the activation function;

[0092] State space: the state is composed of environmental data, electrical data obtained by the data acquisition module, and the functional steady-state deviation degree calculated by the state evaluation module ;

[0093] Action space: the action is defined as the adjustment amount of the inverter control parameter, for example, the fine tuning value of the maximum power point tracking algorithm perturbation step, which is normalized to interval;

[0094] The present embodiment limits the way in which the decision optimization module generates optimal control instructions; this way aims to enable the system to autonomously learn a complex control strategy that takes into account both power generation efficiency and equipment health;

[0095] The process of the decision optimization module generating optimal control instructions is modeled as a deep reinforcement learning task; the inner logic is that:

[0096] The function stability deviation degree calculated by the state evaluation module is taken as a negative penalty item in the preset reward function; this means that any behavior that causes the device to deviate from its healthiest working state will immediately be converted into a negative reward signal through an increased value;

[0097] Based on the preset reward function, a deep reinforcement learning task is used for optimization to generate optimal control instructions for intervention on the target electrical device; in this task, the central processing module acts as an agent that constantly interacts with the physical device or its digital twin model; the agent outputs an action, i.e., the optimal control instruction, at each time step, such as adjusting the MPPT perturbation step size, and the device's state changes after execution; the environment, including the state evaluation module, returns a new state and a reward value calculated based on the reward function; the goal of the agent is to learn an optimal strategy through training that can output actions that maximize the long-term cumulative reward based on the input device state;

[0098] By introducing a deep reinforcement learning framework, a complex control problem is converted into a clear optimization task; this approach frees the system from relying on preset and rigid control rules; the agent can autonomously learn and discover sophisticated control strategies that are difficult for human experts to design through dynamic interaction with the environment, thereby achieving a more intelligent and adaptive balance between power generation income and device health in varying working conditions.

[0099] The preset reward function includes:

[0100] An instantaneous power generation efficiency term representing power generation income;

[0101] A function stability deviation degree representing the health status of the device;

[0102] A grid violation penalty term to ensure grid friendliness;

[0103] Determining the instantaneous power generation efficiency term includes:

[0104] Obtaining the actual output power provided by the data acquisition module; calling the preset digital twin model to calculate the theoretical maximum power under the current environmental conditions; determining the instantaneous power generation efficiency term by calculating the ratio of the actual output power to the theoretical maximum power;

[0105] The determination of the grid violation penalty term includes:

[0106] Comparing the output indicators of the target electrical device with the preset grid specifications;

[0107] When the output index exceeds the limit, the grid violation penalty item is assigned a preset negative value;

[0108] When the output index does not exceed the limit, the grid violation penalty item is assigned a value of zero;

[0109] This embodiment presets the reward function The specific composition of is limited; the design of this reward function is the basis for guiding the deep reinforcement learning agent to learn the desired behavior;

[0110] The default reward function works by combining multiple objectives into a single, dimensionless scalar reward signal using a linear weighted summation approach. In this example, the function consists of three core components:

[0111] The instantaneous power generation efficiency term that represents the power generation revenue;

[0112] The deviation from the functional steady state that characterizes the health status of the equipment;

[0113] Penalties for grid violations to ensure grid friendliness;

[0114] The three together constitute the moment Instant rewards:

[0115] :

[0116] in, For the moment The immediate reward; t is the time; is the instantaneous power generation efficiency calculated by the subsequent steps; subscript g: represents the power grid; subscript t: represents time t; is the steady-state deviation calculated in real time by the state assessment module; Penalties for grid violation; is the dimensionless equilibrium coefficient; Balancing factor The method for determining is to conduct multiple rounds of offline training and evaluation in the digital twin environment to screen out the strategy set that constitutes the Pareto optimal frontier between extending equipment life and sacrificing power generation efficiency. The system operator selects the most appropriate strategy point on this frontier based on its economic model and asset management objectives to determine its final value.

[0117] The operation method of generating Pareto optimal frontier is: To ensure the safety of power grid, Set to a sufficiently large constant by Systematically change within the scope of and For example, take For each ratio, a deep reinforcement learning agent is trained independently until convergence, and the long-term benefits corresponding to each converged strategy, such as the decay inhibition factor and short-term costs, such as regulatory efficiency loss As a point plotted in a two-dimensional coordinate system, the set of points constitutes the discretized Pareto optimal frontier for operators to choose;

[0118] Instantaneous power generation efficiency The determination process is as follows: Get the actual output power provided by the data acquisition module ; Call the preset digital twin model to calculate the theoretical maximum power under current environmental conditions, such as light and temperature The theoretical maximum power is the instantaneous power reference value that can be achieved using the standard MPPT algorithm under current environmental conditions. The instantaneous power generation efficiency term is determined by calculating the ratio of actual output power to theoretical maximum power:

[0119]

[0120] in, It is the instantaneous power generation efficiency; is the actual output power obtained in real time by the data acquisition module; is the theoretical maximum power calculated in real time by the digital twin model; this normalization process makes Becomes a dimensionless value ranging from 0 to 1, ensuring the dimensional consistency of the reward function; subscript m: represents the maximum;

[0121] Penalties for grid violations The determination process is as follows: compare the output indicators of the target electrical equipment, such as voltage, frequency, and harmonic content, with the preset grid specifications; Predefined grid specification Derived from the grid connection technical standards issued by the power sector, the set logic is to serve as a hard constraint to ensure grid security. When any output indicator exceeds the limit, the grid violation penalty item is assigned a preset large negative value.

[0122] To ensure the absolute effectiveness of the penalty, the large negative value should be set to be at least greater than the benefit term in the reward function. The maximum positive value that can be achieved in one time step is an order of magnitude larger; this ensures that any exploration behavior that violates the grid code will immediately receive strong negative feedback, thereby guiding the agent to learn to avoid such dangerous actions in the early stages of policy learning;

[0123] When all output indicators are within the limit, the grid violation penalty item is assigned a value of zero;

[0124] The gain technical effect of this embodiment is:

[0125] Comprehensiveness of objectives: Building a comprehensive consideration of power generation benefits , equipment health and grid security The reward function of three core dimensions ensures the comprehensiveness and practicality of the control strategy;

[0126] Quantification and closed loop: The output of the state assessment module Directly serving as the core negative penalty term in the reward function, it establishes a critical closed-loop feedback path; any behavior that harms health will be immediately punished, guiding the agent to learn to avoid such behavior;

[0127] Constraints and practicality: By incorporating grid-connection regulations into the learning process as hard constraints through grid violation penalties, the generated control strategy is guaranteed to be safe and feasible in engineering.

[0128] This solution achieves a dynamic optimal balance between power generation revenue and equipment health; the decision optimization module uses the functional steady-state deviation as a core negative penalty term in the preset reward function and uses deep reinforcement learning tasks for optimization; the reward function is designed to be ,in, : Instant rewards, : Instantaneous power generation efficiency, : Steady-state deviation, : Penalty items for power grid violations, : Balance coefficient; the construction of this function is logically self-consistent and complete; The term is the instantaneous power generation efficiency term, which is normalized by the ratio of actual output power to theoretical maximum power, ensuring the pursuit of power generation benefits; The term represents the health cost paid to obtain the benefit; That is, the grid violation penalty item ensures that all optimization behaviors are carried out within the boundaries of grid safety regulations; compared with the single, static control objectives of existing technologies, the intelligent agent of this scheme can autonomously learn and execute a set of complex, dynamic optimal control instructions under variable working conditions to maximize long-term cumulative rewards, thereby achieving the Pareto optimality of power generation revenue and equipment health in practice.

[0129] Also includes:

[0130] The performance evaluation module is used to determine the predicted life under the steady-state control strategy composed of optimal control instructions and the predicted life under the standard control strategy based on the preset digital twin model. By calculating the ratio of the two, the decline inhibition factor that characterizes the long-term effectiveness of the steady-state control strategy is obtained;

[0131] The performance evaluation module is also used to:

[0132] determining an actual output power guided by the optimal control instruction under the method;

[0133] determining a theoretical maximum power under current environmental conditions;

[0134] calculating an integral of a difference between the actual output power and the theoretical maximum power within a preset time;

[0135] calculating an integral of the theoretical maximum power within the preset time;

[0136] determining a regulation efficiency loss rate representing a short-term cost of the steady-state control strategy by dividing the integral of the difference by the integral of the theoretical maximum power;

[0137] The system further comprises a performance evaluation module; the purpose of the module is not to participate in real-time closed-loop control, but to serve as an offline evaluation tool for verifying the performance of a control strategy that has been trained, and to macroscopically evaluate its effectiveness from two dimensions of long-term benefit and short-term cost;

[0138] To further illustrate, the performance evaluation module is configured to determine a degradation suppression factor ; Degradation inhibitor ​ The definition of the degradation suppression factor is to intuitively measure the effectiveness of the method in prolonging the service life of the equipment; the determination process is to determine the predicted life under the steady-state control strategy composed of the optimal control instruction based on the aging module in the preset digital twin model , and to determine the predicted life under the standard control strategy under the same external conditions ;

[0139] The standard control strategy here specifically refers to the control strategy that is the only target of maximizing instantaneous power generation by default when the inverter equipment is manufactured by the manufacturer, and is usually the classic perturb and observe or incremental conductance maximum power point tracking algorithm, which does not consider and intervene in the equipment health status during operation;

[0140] By calculating the ratio of the two, the degradation suppression factor representing the long-term benefit of the steady-state control strategy is obtained:

[0141]

[0142] wherein, is the degradation suppression factor; is the predicted life under the steady-state control; is the predicted life under the standard control, both of which are obtained by long-term simulation prediction in the digital twin model;

[0143] To evaluate the short-term cost, the performance evaluation module is further configured to determine a regulation efficiency loss rate The definition of the regulation efficiency loss rate is to quantify the cost of short-term power generation for maintaining the long-term health of the device; its determination process is:

[0144] The actual output power guided by the optimal control instruction under the method , and the theoretical maximum power under the current environmental condition ; Calculate the integral of the difference between the actual output power and the theoretical maximum power, and the integral of the theoretical maximum power within the preset time ; Determine the regulation efficiency loss rate representing the short-term cost of the steady-state control strategy by dividing the difference integral by the integral of the theoretical maximum power:

[0145]

[0146] Wherein, is the efficiency loss rate; is the theoretical maximum power calculated by the digital twin model; is the actual output power measured in the actual operation of the system; is the time; T: the preset integral time period; subscript a: represents actual; : integral symbol; d: differential symbol;

[0147] The gain technical effect of the embodiment is to provide a complete and objective strategy offline evaluation system; through the recession suppression factor and the regulation efficiency loss rate These two complementary indicators, system operators can clearly quantify the long-term benefits of any health management strategy, how many times the life is extended and the short-term cost, how much power generation efficiency is lost; Such quantitative evaluation results provide direct and reliable decision-making basis for operators to choose between different control strategies, or to make a wise trade-off between benefits and costs that meet their business model;

[0148] The scheme provides a complete macroscopic evaluation system for quantitatively evaluating the benefits and costs of strategies; by introducing the recession suppression factor Wherein, : recession suppression factor, : predicted life under steady-state control, : predicted life under standard control and regulation efficiency loss rate Wherein, : efficiency loss rate, is the theoretical maximum power calculated by the digital twin model; is the actual output power measured in the actual operation of the system, : time, long-term benefits of the present solution, life extension and short-term costs, efficiency losses are clearly revealed; for example, a specific strategy can achieve , i.e. the device predicts life extension of 40%, while the cost is only , annual loss of power generation; this quantitative cost-benefit analysis capability enables power plant operators to make data-driven, wise asset management decisions based on their economic models, which is unprecedented in the prior art field;

[0149] In summary, the present solution builds a complete technical system including data acquisition, quantitative evaluation, intelligent decision and closed-loop control, fundamentally solves the inherent contradiction of the prior art that sacrifices long-term health of equipment for short-term power generation, realizes the collaborative management and optimization of the efficiency and health of the electrical equipment of the photovoltaic power station, and has significant technical progress and great application value.

Claims

1. An intelligent performance evaluation and management system for electrical equipment in photovoltaic power plants, characterized by: include: Data acquisition module, used to obtain environmental data, electrical data and operating status data of target electrical equipment in real time; A state assessment module, configured to determine a functional steady-state deviation of the target electrical equipment based on the data acquired by the data acquisition module and in combination with a preset functional steady-state interval; A decision optimization module is used to generate optimal control instructions based on the functional steady-state deviation and a preset reward function; A control execution module is used to execute the optimal control instructions to regulate the operating state of the target electrical equipment; The state assessment module determines the degree of deviation from functional homeostasis, including: Extracting the current value of the stress index from the data acquired by the data acquisition module; calling a preset functional steady-state interval, which is defined by a preset optimal center value of the stress index and a preset safety range of the stress index; calling a preset weight coefficient; and calculating and generating a functional steady-state deviation based on the current value of the stress index, the functional steady-state interval, and the weight coefficient; Determination of the weight coefficient includes: In the pre-set digital twin model, systematic perturbations are performed on each stress indicator. Through accelerated simulation, the negative impact of each stress indicator on the predicted remaining service life of the equipment is quantified. Normalization is performed based on the degree of negative impact to determine the corresponding weight coefficient for each stress indicator. The decision optimization module generates the optimal control instructions, including: The functional steady-state deviation is used as the negative penalty term in the preset reward function. Based on the preset reward function, deep reinforcement learning tasks are used to search for the optimal control instructions for intervening in the target electrical equipment.

2. The intelligent performance evaluation and management system for electrical equipment in photovoltaic power plants according to claim 1, characterized in that: Preset reward functions, including: The instantaneous power generation efficiency term that represents the power generation revenue; The deviation from the functional steady state that characterizes the health status of the equipment; Penalties for grid violations to ensure grid friendliness.

3. The intelligent performance evaluation and management system for electrical equipment in photovoltaic power plants according to claim 2, characterized in that: Determining the instantaneous power generation efficiency item includes: Obtain the actual output power provided by the data acquisition module; call the preset digital twin model to calculate the theoretical maximum power under the current environmental conditions; determine the instantaneous power generation efficiency item by calculating the ratio of the actual output power to the theoretical maximum power.

4. The intelligent performance evaluation and management system for electrical equipment in photovoltaic power plants according to claim 2, characterized in that: The determination of the penalty items for grid violation includes: Comparing the output indicators of the target electrical equipment with the preset grid specifications; When the output index exceeds the limit, the grid violation penalty item is assigned a preset negative value; When the output indicator does not exceed the limit, the grid violation penalty item is assigned a value of zero.

5. The intelligent performance evaluation and management system for electrical equipment in photovoltaic power plants according to claim 1, characterized in that: Also includes: The performance evaluation module is used to determine the predicted life under the steady-state control strategy composed of optimal control instructions based on the preset digital twin model, and to determine the predicted life under the standard control strategy. By calculating the ratio of the two, the decay inhibition factor that characterizes the long-term benefits of the steady-state control strategy is obtained.

6. The intelligent performance evaluation and management system for electrical equipment in photovoltaic power plants according to claim 5, characterized in that: The performance evaluation module is also used to: Determine the actual output power guided by the optimal control command under this method; Determine the theoretical maximum power under current environmental conditions; Calculate the integral of the difference between the actual output power and the theoretical maximum power within the preset time; Calculate the integral of the theoretical maximum power within the preset time; The control efficiency loss rate representing the short-term cost of the steady-state control strategy is determined by dividing the integral of the difference by the integral of the theoretical maximum power.

Images