Networking energy storage inverter state monitoring method based on multi-scale slope dynamic entropy

Abstract

The invention discloses a networking energy storage inverter state monitoring method based on multi-scale slope dynamic entropy, and the method comprises the steps: collecting an output voltage signal of a networking energy storage inverter, and carrying out the preprocessing and phase locking processing, and obtaining an instantaneous phase angle time sequence; performing multi-scale slope calculation and symbolization mapping on the instantaneous phase angle time sequence to generate a discrete symbol sequence; performing phase-space reconstruction and mode extraction on the discrete symbol sequence to obtain a symbol mode set; calculating a multi-scale slope dynamic entropy value based on the occurrence probability of the symbol mode set; and based on the multi-scale slope dynamic entropy, performing no-reference health state evaluation and early warning. According to the scheme, the method is a novel monitoring method which can sensitively capture the consistency change of the dynamic response characteristics of the internal phase angle of the inverter only through the local output signal of the inverter without depending on an external power grid reference signal, and can carry out online early warning on early performance degradation of core functions such as virtual inertia.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of grid-forming energy storage inverter monitoring, and particularly to a grid-forming energy storage inverter state monitoring method based on multi-scale slope dynamic entropy. BACKGROUND

[0002] As a key equipment for building a new power system, the grid-forming energy storage system simulates the operating characteristics of a synchronous generator to provide necessary voltage, frequency support and inertia response (i.e., virtual inertia) for the power grid. The core is that the grid-forming inverter generates a stable output voltage amplitude and phase through an internal control algorithm, thereby autonomously establishing the voltage and frequency of the power grid. The instantaneous phase angle dynamics of the inverter output directly reflects the law of the virtual rotor motion, and is a core state variable for evaluating the quality and consistency of the virtual inertia response.

[0003] Currently, the monitoring and fault diagnosis of the health state of the grid-forming inverter mainly focuses on two aspects: one is the state monitoring of the hardware power components (such as IGBT, DC link capacitor), which mainly uses temperature, voltage, current and other physical signal monitoring combined with model or data-driven methods; the other is the stability analysis of the external characteristics of the grid-connected system, which usually relies on grid-side frequency measurement, power measurement and other signals, and evaluates the performance by comparing the relationship between the inverter response and the grid disturbance.

[0004] However, the above existing technologies have obvious limitations: Firstly, for the internal health state, the traditional method is difficult to effectively capture the slow degradation of the control loop performance. The "health" of the grid-forming inverter is not only that the hardware is fault-free, but also that the software control algorithm can continuously and consistently output the dynamic phase angle response as expected. The performance degradation caused by control parameter drift, software anomaly or component aging may still maintain normal amplitude of the output electrical signal before the hardware completely fails, but the dynamic characteristics of the phase angle change have already appeared subtle distortion, and the traditional amplitude-based monitoring method is extremely insensitive to this.

[0005] Secondly, for the external performance evaluation, the existing method relies heavily on the grid reference signal. Whether it is based on frequency response to evaluate virtual inertia or based on power instruction tracking to evaluate damping characteristics, it needs a reference signal (such as frequency change curve) from the grid that is considered to be "normal" for comparison. This can be achieved when operating in grid-connected mode, but once the system enters island operation or weak grid operating conditions, the grid reference signal itself may be distorted or fluctuate sharply, making such methods ineffective and unable to achieve truly autonomous and reference-free monitoring.

[0006] In addition, the existing method lacks the early warning ability of early and weak faults. Slowly occurring performance degradation only shows a slight increase in the complexity of dynamic response nonlinearity at the beginning, and traditional time domain or frequency domain analysis methods are difficult to extract such predictive features from strong background noise, often discovered when the fault has developed into a hard fault, causing protection action, and cannot achieve predictive maintenance. SUMMARY

[0007] The present application provides a grid-connected energy storage inverter state monitoring method based on multi-scale slope dynamic entropy, which can solve the above problems existing in the prior art. The present application is a new monitoring method that can sensitively capture the consistency change of the internal phase angle dynamic response characteristics of the inverter itself without relying on external grid reference signals, and can perform online early warning on the early performance degradation of virtual inertia and other core functions.

[0008] The technical scheme is a grid-connected energy storage inverter state monitoring method based on multi-scale slope dynamic entropy, which comprises the following steps: Collecting the output voltage signal of the grid-connected energy storage inverter, and obtaining the instantaneous phase angle time sequence through preprocessing and phase-locked processing; Performing multi-scale slope calculation and symbolic mapping on the instantaneous phase angle time sequence to generate a discrete symbol sequence; Performing phase space reconstruction and pattern extraction on the discrete symbol sequence to obtain a symbol pattern set; Calculating the multi-scale slope dynamic entropy value based on the occurrence probability of the symbol pattern set; Performing no-reference health state evaluation and early warning based on the multi-scale slope dynamic entropy value.

[0009] According to a further improvement of the present application, the collecting of the output voltage signal of the grid-connected energy storage inverter, and the obtaining of the instantaneous phase angle time sequence through preprocessing and phase-locked processing, comprises: Collecting the single-phase voltage instantaneous analog signal of the inverter output end through a voltage sensor; Performing analog-to-digital conversion on the instantaneous analog signal to obtain a digital voltage signal; Performing band-pass filtering processing on the digital voltage signal to obtain a filtered voltage signal; Performing software phase-locked loop processing on the filtered voltage signal to calculate the instantaneous phase angle time sequence.

[0010] According to a further improvement of the present application, the software phase-locked loop processing on the filtered voltage signal to calculate the instantaneous phase angle time sequence comprises: Performing orthogonal transformation on the filtered voltage signal to generate two-phase orthogonal voltage components in the alpha-beta coordinate system; calculating phase errors of the two-phase quadrature voltage components and internally generated quadrature signals by a phase detector; inputting the phase errors into a loop filter for filtering to obtain a control voltage; generating a rotation angle of a synchronous rotating coordinate system according to the control voltage by a voltage-controlled oscillator; integrating the rotation angle to obtain a time series of instantaneous phase angles.

[0011] According to a further improvement of the present application, the multi-scale slope calculation and symbolic mapping of the time series of instantaneous phase angles to generate a discrete symbol sequence comprises: setting one or more time scale factors; for each scale factor, calculating a differential slope value of the phase angle sequence at the corresponding scale; setting an angle threshold value and comparing the differential slope value with the angle threshold value; according to the comparison result, mapping consecutive slope values into discrete symbols to generate a symbol sequence composed of -1, 0, +1; the method is as follows: setting a positive angle threshold value ε+ and a negative angle threshold value ε-; if the slope value is greater than ε+, then mapping it into symbol +1; if the slope value is less than ε-, then mapping it into symbol -1; if the slope value is between ε- and ε+, then mapping it into symbol 0.

[0012] According to a further improvement of the present application, the calculation of a differential slope value of the phase angle sequence at the corresponding scale for each scale factor comprises: reading a current time scale factor τ and calculating a time offset; for each data point θ(t) in the phase angle time sequence, obtaining a data point θ(t+τ) after time offset; calculating a differential value between adjacent data points: Δθ = θ(t+τ) - θ(t); calculating a differential slope value: slope = Δθ / (τ × Δt), wherein Δt is a sampling time interval; repeating the above calculation for all data points to obtain a complete differential slope sequence at the current scale.

[0013] According to a further improvement of the present application, the phase space reconstruction and pattern extraction of the discrete symbol sequence to obtain a symbol pattern set comprises: setting an embedding dimension m and a time delay parameter L; reconstructing the symbol sequence in a phase space according to the embedding dimension and the time delay parameter to generate symbol pattern vectors; traversing all symbol pattern vectors to identify and record unique symbol patterns to form a symbol pattern set.

[0014] According to a further improvement of the present application, the phase space reconstruction of the symbol sequence according to the embedding dimension and time delay to generate the symbol pattern vector comprises: Starting from the first symbol of the symbol sequence, sequentially select m consecutive symbols, and the time interval between adjacent symbols is L; Combine the selected m symbols in time sequence to form an m-dimensional symbol pattern vector; Slide the time window and repeat the above selection and combination process until the entire symbol sequence is traversed; All generated symbol pattern vectors form a symbol pattern matrix.

[0015] According to a further improvement of the present application, the calculation of the multiscale slope dynamic entropy value based on the occurrence probability of the symbol pattern set comprises: Count the number of occurrences of each unique symbol pattern in the reconstructed phase space; Calculate the occurrence probability of each symbol pattern, including: count the total number of symbol patterns N in the reconstructed phase space; count the number of occurrences n_i of each unique symbol pattern; calculate the occurrence probability of each symbol pattern: P_i = n_i / N; According to the information entropy formula, calculate the slope dynamic entropy value based on the occurrence probability of all symbol patterns.

[0016] According to a further improvement of the present application, the no-reference health state evaluation and early warning based on the multiscale slope dynamic entropy value comprises: Pre-record the baseline entropy value curve under different operating conditions under the inverter health state; Compare the real-time calculated multiscale entropy value with the baseline entropy value curve under the corresponding operating condition; If the real-time entropy value is continuously higher than the baseline entropy value by more than a preset tolerance, it is determined that the virtual inertia consistency is deteriorated; If the real-time entropy value shows a trend of rising, an early performance degradation warning signal is generated.

[0017] According to a further improvement of the present application, the process of comparing the real-time calculated multiscale entropy value with the baseline entropy value curve under the corresponding operating condition comprises: Calculate the deviation ΔSE of the real-time entropy value and the baseline entropy value; Standardize the deviation: ΔSE_norm = (ΔSE - μ) / σ, where μ and σ are the mean and standard deviation of the baseline deviation; Based on whether the standardized deviation exceeds the dynamic threshold, trigger different levels of warning.

[0018] Beneficial effects: This invention enables autonomous health monitoring of grid-connected inverters without relying on grid reference signals, and can work effectively under various operating conditions such as islanding and weak grids; by analyzing the dynamic entropy of the phase angle slope, this method has extremely high detection sensitivity for early and slight degradation of control performance, and can achieve predictive maintenance; this method has high computational efficiency, requires only a single-phase voltage signal, is easy to integrate into existing inverter control systems, and has strong engineering applicability. Attached Figure Description

[0019] Figure 1 This is the overall flowchart of the present invention.

[0020] Figure 2 This is a flowchart of the multi-scale slope dynamic entropy calculation of the present invention.

[0021] Figure 3 This is a flowchart of the reference-free health assessment of the present invention.

[0022] Figure 4 This is a schematic diagram of the application scenario of the system of the present invention. Detailed Implementation

[0023] To better illustrate the technical solution of the present invention, the following will be combined with... Figures 1 to 4 The present invention will be described in detail with reference to specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of this application can be combined with each other. Example

[0024] like Figures 1 to 4 As shown, the method for monitoring the state of grid-connected energy storage inverters based on multi-scale slope dynamic entropy includes the following steps: The output voltage signal of the grid-type energy storage inverter is collected, and the instantaneous phase angle time series is obtained after preprocessing and phase-locked loop processing. The instantaneous phase angle time series is subjected to multi-scale slope calculation and symbolic mapping to generate a discrete symbolic sequence; The discrete symbol sequence is reconstructed in phase space and the pattern is extracted to obtain a symbol pattern set; Based on the occurrence probability of the symbol pattern set, calculate the multi-scale slope dynamic entropy value; Based on the multi-scale slope dynamic entropy value, a no-reference health status assessment and early warning are performed.

[0025] According to a further improvement of the present invention, the acquisition of the output voltage signal of the grid-type energy storage inverter, after preprocessing and phase-locked loop processing to obtain the instantaneous phase angle time series, includes: The instantaneous analog signal of the single-phase voltage at the output of the inverter is acquired by a voltage sensor; The instantaneous analog signal is converted from analog to digital to obtain a digital voltage signal; The digital voltage signal is band-pass filtered to obtain a filtered voltage signal. The filtered voltage signal is processed by a software phase-locked loop to obtain an instantaneous phase angle time series.

[0026] According to a further improvement of the application, the processing of the filtered voltage signal by a software phase-locked loop to obtain an instantaneous phase angle time series comprises: The filtered voltage signal is subjected to a quadrature transformation to generate two-phase quadrature voltage components in an α-β coordinate system. A phase detector is used to calculate the phase error between the two-phase quadrature voltage components and internally generated quadrature signals. The phase error is input into a loop filter for filtering to obtain a control voltage. A voltage-controlled oscillator is used to generate the rotation angle of a synchronous rotating coordinate system according to the control voltage. The rotation angle is subjected to an integral operation to obtain an instantaneous phase angle time series.

[0027] According to a further improvement of the application, the multi-scale slope calculation and symbolic mapping of the instantaneous phase angle time series to generate a discrete symbol sequence comprises: One or more time scale factors are set. For each scale factor, the differential slope value of the phase angle sequence at the corresponding scale is calculated. An angle threshold is set, and the differential slope value is compared with the angle threshold. According to the comparison result, the continuous slope value is mapped to a discrete symbol to generate a symbol sequence composed of -1, 0, and +1. The method is as follows: a positive angle threshold ε+ and a negative angle threshold ε- are set; if the slope value is greater than ε+, the symbol +1 is mapped; if the slope value is less than ε-, the symbol -1 is mapped; if the slope value is between ε- and ε+, the symbol 0 is mapped.

[0028] According to a further improvement of the application, the calculation of the differential slope value of the phase angle sequence at the corresponding scale for each scale factor comprises: The current time scale factor τ is read, and a time offset is calculated. For each data point θ(t) in the phase angle time series, a time-offset data point θ(t+τ) is obtained. The differential value between adjacent data points is calculated: Δθ = θ(t+τ) - θ(t). The differential slope value is calculated: slope = Δθ / (τ × Δt), where Δt is the sampling time interval. The above calculation is repeated for all data points to obtain a complete differential slope sequence at the current scale.

[0029] According to a further improvement of the present application, the phase space reconstruction and pattern extraction on the discrete symbol sequence to obtain a symbol pattern set comprises: setting an embedding dimension m and a time delay parameter L; reconstructing the symbol sequence in phase space according to the embedding dimension and the time delay parameter to generate a symbol pattern vector; traversing all the symbol pattern vectors to identify and record unique symbol patterns to form a symbol pattern set.

[0030] According to a further improvement of the present application, the phase space reconstruction and pattern extraction on the discrete symbol sequence to obtain a symbol pattern set comprises: starting from the first symbol of the symbol sequence, sequentially selecting m consecutive symbols with a time interval L between adjacent symbols; combining the selected m symbols in time sequence to form an m-dimensional symbol pattern vector; sliding the time window to repeat the above selection and combination process until the entire symbol sequence is traversed; composing all the generated symbol pattern vectors into a symbol pattern matrix.

[0031] According to a further improvement of the present application, the calculation of the multi-scale slope dynamic entropy value based on the occurrence probability of the symbol pattern set comprises: counting the number of occurrences of each unique symbol pattern in the reconstructed phase space; calculating the occurrence probability of each symbol pattern, including: counting the total number N of symbol patterns in the reconstructed phase space; counting the number n_i of occurrences of each unique symbol pattern; and calculating the occurrence probability P_i of each symbol pattern: P_i = n_i / N; calculating the slope dynamic entropy value based on the occurrence probability of all symbol patterns according to the information entropy formula.

[0032] According to a further improvement of the present application, the no-reference health state evaluation and early warning based on the multi-scale slope dynamic entropy value comprises: previously recording a baseline entropy value curve under different operating conditions under the health state of the inverter; comparing the real-time calculated multi-scale entropy value with the baseline entropy value curve under the corresponding operating condition; if the real-time entropy value continuously exceeds the baseline entropy value by more than a preset tolerance, it is determined that the virtual inertia consistency is deteriorated; if the real-time entropy value shows a trend of rising, an early performance degradation warning signal is generated.

[0033] According to a further improvement of the present application, the process of comparing the real-time calculated multi-scale entropy value with the baseline entropy value curve under the corresponding operating condition comprises: Calculate the deviation of real-time entropy value and reference entropy value ΔSE; Standardize the deviation: ΔSE_norm = (ΔSE - μ) / σ, where μ and σ are the mean and standard deviation of reference deviation; Trigger different levels of early warning based on whether the standardized deviation exceeds the dynamic threshold. Embodiments

[0034] The network-constructed energy storage inverter state monitoring method based on multi-scale slope dynamic entropy discards the traditional external reference comparison idea and creatively proposes that the "health" of the network-constructed inverter is the consistency of its dynamic response mode. Any internal component aging or control abnormality will cause slight and nonlinear distortion of the slope mode of its dynamic response (especially the phase angle change), and this distortion is manifested as an increase in complexity (i.e., a decrease in regularity) in the early stage.

[0035] We use multi-scale slope symbolic dynamic entropy to directly analyze the phase angle time series of the inverter output voltage, convert the phase angle dynamic response generated locally and carrying virtual inertia information into a quantifiable "consistency entropy value". Abnormal changes in the entropy value are the earliest signs of internal performance degradation.

[0036] The following is the detailed data processing process of the online operation of the system: Step one: signal acquisition and preprocessing Input signal: high-speed sampling (e.g., 10 kHz) of the network-constructed energy storage inverter output A-phase voltage instantaneous value u(t). (Note: selecting a single-phase voltage can reflect the phase angle dynamic without the need for three-phase voltage, simplifying the system) Preprocessing: band-pass filtering of u(t) to filter out high-frequency switching noise and extremely low-frequency drift, retaining the dynamic near the fundamental frequency (e.g., 45-55 Hz).

[0037] Core conversion: real-time calculation of the instantaneous phase angle θ(t) of the voltage through a software phase-locked loop (Software PLL). θ(t) is the core internal state variable of network control, directly reflecting the position of the virtual rotor angle, and its dynamic characteristics contain all the response information of virtual inertia, damping, etc. At this point, we have converted the problem to the analysis of the time series of θ(t).

[0038] Step two: multi-scale phase angle slope calculation and symbolization Compute the slope angle: For the phase angle sequence θ(t), compute the slope (i.e. the instantaneous angular velocity deviation) between adjacent sampling points by: slope(i) = [θ(i+1) - θ(i)] / Δt (Δt is the sampling interval). To further capture the dynamics, one can compute the slope across multiple time scales (multi-scale analysis), e.g. scale factor τ (e.g. τ = 1, 2, 4,...), compute the slope between θ(t) and θ(t+τ).

[0039] Symbolize: Set a small angle threshold ε (e.g. 0.001 radian), which is not used to distinguish between positive and negative, but to filter out noise and small stationary fluctuations. If slope(i) > +ε, symbolize as +1 (indicating the phase angle is "accelerating forward"); if slope(i) < -ε, symbolize as -1 (indicating the phase angle is "decelerating backward"); if -ε <= slope(i) <= +ε, symbolize as 0 (indicating the phase angle is in a "relatively stationary" state). The innovation here is the introduction of the "0" symbol, which accurately depicts the micro-dynamics of the networked system near the equilibrium point, which is the most sensitive interval for fault initiation.

[0040] Step 3: Construct the symbol sequence and phase space reconstruction Form the symbol sequence: Symbolize the slope calculation results within a time window (e.g. several seconds of data containing a once-frequency disturbance event) to obtain a symbol sequence S = {s1, s2, s3,..., sN} composed of {-1, 0, +1}.

[0041] Phase space reconstruction: Set the embedding dimension m (pattern length, e.g. 4) and time delay L (e.g. 1). Reconstruct the symbol sequence S into a matrix, with each row being a symbol pattern: Pattern1: [s1, s1+L, s1+2L,..., s1+(m-1)L]; Pattern2: [s2, s2+L, s2+2L,..., s2+(m-1)L];... These patterns represent specific slope change patterns in the phase angle dynamic evolution over a period of time.

[0042] Step 4: Calculate the multi-scale slope dynamic entropy Statistical pattern frequency: Count the number of unique symbol patterns in the reconstructed phase space and calculate their probability P(pattern_i).

[0043] Calculate the entropy value: Calculate the slope entropy SE according to the information entropy formula: SE = -∑ [P(pattern_i) * log_base(P(pattern_i))] (sum over all unique patterns); where log_base can be chosen as needed (such as natural logarithm).

[0044] Multi-scale analysis: Repeat steps two to four for different time scale factors τ to obtain a set of entropy values SE(τ), forming an entropy-scale curve. The curve of a healthy system should be smooth and predictable, while the curve of a degraded system may show abnormal fluctuations or overall rise.

[0045] Step five: No-reference health assessment and early warning Baseline establishment: When the system is brand new and debugged correctly, apply a small perturbation (or use the inherent perturbation of the power grid) at different power levels to record a series of baseline entropy curves.

[0046] Online monitoring and evaluation: Calculate the real-time entropy value online and compare it with the baseline curve under the corresponding operating condition.

[0047] Consistency evaluation: If the current entropy value is consistently and significantly higher than the baseline entropy value (e.g., more than 3 standard deviations from the baseline), it indicates that the regularity of the phase angle dynamics is decreasing and the complexity is increasing, and the consistency of the virtual inertia response is deteriorating.

[0048] Early warning: The trend of entropy value rising (even if it does not exceed the threshold) can send an early warning, prompting maintenance personnel to pay attention to the internal state of the inverter (such as DC link capacitor value, gate drive characteristics, control parameters, etc.), and achieving predictive maintenance.

[0049] Standardized output (optional): The entropy value can be standardized to the interval [0, 1], where 0 represents perfect consistency and 1 represents complete randomness, and the "health degree" can be intuitively displayed.

[0050] The advantages and innovations of this scheme are as follows: Absolute innovation: The concept and method of "no-reference autonomous evaluation of virtual inertia consistency" are first proposed, which does not rely on any signals from the power grid side, and is truly "self-aware" intelligent.

[0051] Depth and professionalism: Directly hitting the core "virtual inertia" nature of grid-connected energy storage, it abstracts it as a "complexity of phase angle slope sequence" problem from the unique perspective of nonlinear dynamics, and the diagnosis object goes from "components" to "control performance" itself.

[0052] High sensitivity and early warning: The symbolic method is extremely sensitive to small and stable fluctuations (0 symbols) and small abnormal changes, and the entropy value can capture early and slow performance drift that traditional amplitude and spectral analysis cannot detect.

[0053] Strong applicability: suitable for various operation modes such as grid-connected and island, especially obvious advantages in micro-grid and weak grid scenarios.

[0054] Engineering friendliness: moderate calculation amount in the processing process, only single-phase voltage sensor and PLL calculation result on the existing control board are needed, and it is easy to realize through software upgrade on the existing equipment. Embodiment

[0055] The networked energy storage inverter state monitoring method based on multi-scale slope dynamic entropy comprises the following steps: S1: signal acquisition and phase angle information extraction process S1.1: acquire original voltage data. Through the voltage sensor and the analog-to-digital converter, the single-phase voltage instantaneous value signal u(t) of the output end of the networked energy storage inverter is acquired.

[0056] S1.2: signal preprocessing. The acquired voltage signal u(t) is subjected to digital band-pass filtering processing to filter out high-frequency switching noise and low-frequency drift, and a pure filtered voltage signal u_f(t) is obtained.

[0057] S1.3: calculate the instantaneous phase angle. The filtered voltage signal u_f(t) is input into the software phase-locked loop for instantaneous phase locking and calculation processing to obtain the instantaneous phase angle time series data θ(t) which can represent the virtual rotor angle position.

[0058] S2: multi-scale phase angle slope calculation and symbolization process S2.1: calculate the slope of adjacent points. For the phase angle sequence θ(t), according to one or more preset time scale factors τ, the difference slope value slope_τ(t) between the phase angle data points under the scale τ is calculated.

[0059] S2.2: perform slope symbolization mapping. The continuous slope value slope_τ(t) calculated is compared with a preset minimum angle threshold ε, and symbolization mapping processing is performed according to the comparison result: if slope_τ(t)>+ε, it is mapped to symbol +1; if slope_τ(t)<-ε, it is mapped to symbol -1; if -ε<= slope_τ(t)<= +ε, it is mapped to symbol 0. Through this step, a discrete symbol sequence S_τ composed of {-1, 0, +1} is obtained.

[0060] S3: symbol sequence reconstruction and pattern generation process S3.1: reconstruct the phase space. Set the embedding dimension m and the time delay L, and perform phase space reconstruction processing on the symbol sequence S_τ to generate a symbol pattern matrix composed of continuous symbols.

[0061] S3.2: Generate unique patterns. Traverse the symbol pattern matrix, identify and record all unique symbol patterns of length m.

[0062] S4: Symbol pattern probability statistics and entropy calculation process S4.1: Count the frequency of pattern occurrence. Count the number of times each unique symbol pattern appears in the symbol pattern matrix.

[0063] S4.2: Calculate the probability of pattern occurrence. According to the frequency of each pattern, calculate its probability value P(i) in the entire pattern sequence.

[0064] S4.3: Calculate the slope dynamic entropy. Substitute the occurrence probability P(i) of each unique pattern into the information entropy calculation formula to calculate the entropy value, and obtain the slope dynamic entropy value SE(τ) under the current time window and current scale τ.

[0065] S5: Multi-scale analysis and health status evaluation process S5.1: Construct the entropy-scale curve. Change the time scale factor τ, repeat steps S2 to S4, obtain a series of entropy values SE(τ) under different scales, and combine to form a multi-scale entropy-scale curve.

[0066] S5.2: Perform no-reference health assessment. Compare and analyze the real-time calculated entropy value or entropy-scale curve with the baseline entropy curve established in the healthy state and pre-stored in the database: If the real-time entropy value or curve shape deviates significantly from the baseline curve (such as the entropy value is consistently high), a virtual inertia consistency degradation evaluation conclusion is generated; If the entropy value shows a trend of rising, an early performance degradation warning signal is generated.

[0067] S5.3: Output evaluation results. Transmit the evaluation conclusion and warning signal to the local human-machine interface or remote monitoring center for result display and alarm notification.

[0068] This scheme deeply integrates advanced nonlinear time series analysis methods with the core challenges of network-type energy storage, creating a new and original monitoring paradigm, providing a powerful tool for solving the inherent state evaluation problem, and is expected to promote network-type energy storage from "function implementation" to "high reliability, perceptible" new stage.

[0069] The preferred embodiments of the present application are described in detail above, but the present application is not limited to the specific details of the above-described embodiments, and various equivalent transformations of the technical solutions of the present application can be made within the technical concept of the present application, and these equivalent transformations all belong to the protection scope of the present application.

Claims

1. A networked energy storage inverter condition monitoring method based on multi-scale slope dynamic entropy, characterized in that, The method comprises the following steps: Collecting the output voltage signal of the grid-connected energy storage inverter, and obtaining the instantaneous phase angle time sequence through preprocessing and phase-locked processing; Performing multi-scale slope calculation and symbolic mapping on the instantaneous phase angle time sequence to generate a discrete symbol sequence; Performing phase space reconstruction and pattern extraction on the discrete symbol sequence to obtain a symbol pattern set; Based on the occurrence probability of the symbol pattern set, the multi-scale slope dynamic entropy value is calculated; Based on the multi-scale slope dynamic entropy value, the no-reference health state evaluation and early warning are performed.

2. The method of claim 1, wherein, The method for collecting the output voltage signal of the grid-connected energy storage inverter, and obtaining the instantaneous phase angle time sequence through preprocessing and phase-locked processing comprises the following steps: Collecting the single-phase voltage instantaneous analog signal of the output end of the inverter through a voltage sensor; Performing analog-to-digital conversion on the instantaneous analog signal to obtain a digital voltage signal; Performing band-pass filtering on the digital voltage signal to obtain a filtered voltage signal; Performing software phase-locked loop processing on the filtered voltage signal to obtain the instantaneous phase angle time sequence.

3. The method of claim 2, wherein, The method for performing software phase-locked loop processing on the filtered voltage signal to obtain the instantaneous phase angle time sequence comprises the following steps: Performing quadrature transformation on the filtered voltage signal to generate two-phase quadrature voltage components in the α-β coordinate system; Using a phase detector to calculate the phase error of the two-phase quadrature voltage components and the internally generated quadrature signal; Inputting the phase error into a loop filter for filtering to obtain a control voltage; Using a voltage-controlled oscillator to generate the rotation angle of the synchronous rotating coordinate system according to the control voltage; Performing integral operation on the rotation angle to obtain the instantaneous phase angle time sequence.

4. The method of claim 1, wherein, The method for performing multi-scale slope calculation and symbolic mapping on the instantaneous phase angle time sequence to generate a discrete symbol sequence comprises the following steps: Setting one or more time scale factors; For each scale factor, calculating the difference slope value of the phase angle sequence at the corresponding scale; Setting an angle threshold value, and comparing the difference slope value with the angle threshold value; According to the comparison result, the continuous slope value is mapped to a discrete symbol to generate a symbol sequence composed of -1, 0 and +1; the method is as follows: setting a positive angle threshold value ε+ and a negative angle threshold value ε-; if the slope value is greater than ε+, the symbol +1 is mapped; if the slope value is less than ε-, the symbol -1 is mapped; if the slope value is between ε- and ε+, the symbol 0 is mapped.

5. The method of claim 4, wherein, The method for calculating the difference slope value of the phase angle sequence at the corresponding scale for each scale factor comprises the following steps: Reading the current time scale factor τ and calculating the time offset; For each data point θ(t) in the phase angle time sequence, obtaining the data point θ(t+τ) after time offset; Calculating the difference value between adjacent data points: Δθ = θ(t+τ) - θ(t); Calculating the difference slope value: slope = Δθ / (τ × Δt), wherein Δt is the sampling time interval; Repeating the above calculation for all data points to obtain the complete difference slope sequence at the current scale.

6. The method of claim 1, wherein, The method for performing phase space reconstruction and pattern extraction on the discrete symbol sequence to obtain a symbol pattern set comprises the following steps: Setting the embedding dimension m and the time delay parameter L; Reconstruct the phase space according to the embedding dimension and the time delay parameter, and generate a symbol pattern vector from the symbol sequence; Iterate all symbol pattern vectors, identify and record unique symbol patterns, and form a symbol pattern set.

7. The method of claim 6, wherein, The step of reconstructing the phase space according to the embedding dimension and the time delay parameter, and generating a symbol pattern vector from the symbol sequence, comprises: Start from the first symbol of the symbol sequence, and sequentially select m consecutive symbols, with a time interval L between adjacent symbols; Combine the selected m symbols in time sequence to form an m-dimensional symbol pattern vector; Slide the time window and repeat the selection and combination process until the entire symbol sequence is traversed; Combine all generated symbol pattern vectors to form a symbol pattern matrix.

8. The method of claim 1, wherein, The step of calculating the multiscale slope dynamic entropy value based on the occurrence probability of the symbol pattern set, comprises: Count the number of occurrences of each unique symbol pattern in the reconstructed phase space; Calculate the occurrence probability of each symbol pattern, including: counting the total number of symbol patterns N in the reconstructed phase space; counting the number of occurrences n_i of each unique symbol pattern; calculating the occurrence probability of each symbol pattern: P_i = n_i / N; Calculate the slope dynamic entropy value based on the occurrence probability of all symbol patterns according to the information entropy formula.

9. The method of claim 1, wherein, The step of performing no-reference health state evaluation and early warning based on the multiscale slope dynamic entropy value, comprises: Pre-record the baseline entropy value curve under different operating conditions under the inverter health state; Compare the real-time multiscale entropy value with the baseline entropy value curve under the corresponding operating condition; If the real-time entropy value is continuously higher than the baseline entropy value by more than a preset tolerance, it is determined that the virtual inertia consistency is deteriorating; If the real-time entropy value shows a trend of rising, an early performance degradation warning signal is generated.

10. The method of claim 9, wherein, The process of comparing the real-time multiscale entropy value with the baseline entropy value curve under the corresponding operating condition, comprises: Calculate the deviation ΔSE of the real-time entropy value from the baseline entropy value; Standardize the deviation: ΔSE_norm = (ΔSE - μ) / σ, where μ and σ are the mean and standard deviation of the baseline deviation; Trigger different levels of warning based on whether the standardized deviation exceeds the dynamic threshold.

Images