Load time sequence feature considering energy consumption curve local differential privacy protection method

Abstract

The application discloses a kind of energy consumption curve local difference privacy protection methods considering load timing characteristics, comprising: extracting user energy consumption curve features by time series symbol aggregation method, forming a symbolized energy consumption sequence;Using variable window random mapping mechanism to disturb symbol sequence timing, make the feature site of original energy consumption curve change to protect the energy consumption privacy of user;Subsequently, the disturbed symbol sequence is numerized as a new energy consumption curve, relies on energy storage to realize the conversion of original energy consumption curve to new energy consumption curve, the statistical characteristics of the new energy consumption curve formed can be retained under the premise of protecting the load timing privacy of user.The real random noise introduced by energy storage device guarantees the privacy and security of user load information, masks the fluctuation characteristics of user power consumption behavior, and avoids the leakage of privacy information existing in energy consumption timing.

Description

Technical Field

[0001] This invention relates to the field of electrical engineering, and more particularly to a localized differential privacy protection method for energy consumption curves that takes into account load timing characteristics. Background Technology

[0002] Smart meters serve as the data collection portal for new power systems. The fine-grained data they collect is fundamental to the stable operation of the system and also facilitates power grid companies in providing high-quality and efficient services to users based on data characteristics. User energy consumption curves, measured by smart meters, are a crucial representation of user electricity usage. However, with the development of technologies such as non-intrusive load monitoring, data attackers can leverage large amounts of monitored power data to infer equipment operating status, achieve detailed descriptions of user behavior, fine-grained user classification, and high-confidence load forecasts. For example, attackers can use non-intrusive load monitoring technology to accurately predict the sequence of a user's home use of appliances such as electric heaters, blenders, and ovens. Over time, continuous monitoring can further infer private information such as the number of family members, family work schedules, and even health status. The leakage of this personal privacy information poses risks to users' lives, leading to economic losses and even crimes such as theft.

[0003] User energy consumption curves are composed of load data with strong time-series characteristics. Sensitive information exists not only in the numerical changes of electricity load but also hidden in the user's energy consumption time series. The energy consumption time series directly reflects users' electricity usage habits, exhibiting seasonality and regularity. Attackers can infer users' daily behavior and electricity preferences based on the load time series characteristics in the energy consumption curve. However, current privacy protection methods for time series data mostly focus on the numerical values ​​themselves. For example, most differential privacy protection methods introduce independent and identically distributed noise to ensure that the numerical information of the time series cannot be recovered with high confidence. Once an attacker observes and analyzes the fluctuations in the energy consumption curve, users still face the risk of privacy leakage. Furthermore, using randomized noise to protect the energy consumption curve inevitably introduces spikes, leading to a certain degree of data distortion and affecting data usability. The localized differential privacy concept achieves privacy protection through random perturbation, avoiding the drawbacks of traditional differential privacy while inheriting its advantages. If the original time series of the user's energy consumption curve can be disrupted, preventing it from being recovered with high confidence by an attacker, sensitive information about the user's electricity consumption can be effectively concealed. Therefore, this study aims to address the problem that current methods for protecting the privacy of energy consumption curves ignore the strong temporal sequence of load data, which makes it impossible to mask the fluctuating characteristics of users' electricity consumption behavior. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a localized differential privacy protection method for energy consumption curves that considers load timing characteristics, based on the concept of localized differential privacy that takes into account timing. This solves the problem that current privacy protection methods ignore the timing characteristics of electricity consumption curves, which leads to an inability to effectively mask electricity consumption timing. At the same time, it provides a reference for determining household energy storage parameters.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A localized differential privacy protection method for energy consumption curves that considers load timing characteristics includes the following steps:

[0007] Step 1: Symbolize the original user energy consumption curve using the time series symbol aggregation method, transforming the user energy consumption curve from a time series into a discrete character sequence. Each character in the sequence represents a certain load time series feature, thereby effectively extracting the variation features in the energy consumption curve and providing a basis for achieving privacy protection of load time series features.

[0008] Step 2: Using the variable window random mapping mechanism, the symbolic original user energy consumption curve is transformed into a perturbed symbol sequence. The mapping probability of the sequence site with a time window span of k is determined. The symbolic sequence is temporally perturbed by the randomized mapping of the sequence site, and transformed into a perturbed symbol sequence. The symbol sequence is then numerically converted into a new user energy consumption curve to ensure that the sequence can achieve localized differential privacy protection of the original sequence.

[0009] Step 3: Repeat steps 1 and 2 to process the user's historical data for one month to obtain a set of minimum energy storage capacity parameters and rate parameters that can meet all mapping requirements. Repeat this process multiple times, and then select the parameters that meet the confidence threshold from these parameters as the energy storage device parameters required by the user.

[0010] Step 4: Use the charging and discharging behavior of the energy storage device required by the user, as determined in Step 3, to transform the original energy consumption curve into the disturbed energy consumption curve; when the required charging and discharging amount exceeds the parameter limit, the maximum energy that can actually be provided is taken as the final charging and discharging amount of the stored energy.

[0011] Furthermore, step one includes the following sub-steps:

[0012] (a) Normalize the energy consumption curve to transform the numerical range in the time series into [0,1], thereby reducing the impact of adverse data on the series. The normalization method is as follows:

[0013]

[0014] In the formula: The value is the normalized value; The first time series One value; The maximum value in the sequence; The minimum value of the sequence;

[0015] (b) The energy consumption curve is approximated by load segmentation aggregation, and the amplitude of the decrease in the same time interval is replaced by the average value. The calculation method is as follows:

[0016]

[0017] In the formula: For the first The average value of the data segment; For the first The time points of each load data point; An index for normalized load data; This is an index of the load data after approximate transformation through load segmentation and aggregation;

[0018] (c) The energy consumption curve is symbolized as a discrete string. That is, the energy consumption curve after load segmentation and aggregation approximation is divided into intervals according to the amplitude. Each character represents a certain amplitude range. The relationship between the converted amplitude and the character is as follows:

[0019]

[0020] In the formula: The i-th character in the character list; The character in the preset p-th segment has a numerical range of 1. .

[0021] Furthermore, step two specifically includes the following sub-steps:

[0022] (a) Assume that the original energy consumption curve forms an original symbol sequence after being processed by the time series symbol aggregation method. At the same time, there exists a mapping sequence with the same length as the symbol sequence. In the initial stage, all points in the mapping sequence are empty, and a time span is... The window is placed at the beginning of both the original sequence and the mapped sequence;

[0023] (b) Initial random mapping, the original sequence subsequence located within the window First site Mapped to a certain probability Window subsequence of a sequence At any site; subsequently, the mapped sites in the window subsequence of the original sequence are discarded, the window is shifted forward to add new sites and form a new window subsequence, and the number of sequence sites in the window remains unchanged. Mapped sites in the window subsequence of the mapped sequence are discarded, and the window moves one position to the right, while the number of sites in the window remains the same. ;

[0024] (c) In each subsequent random mapping, referring to step (b), a site in the window subsequence of the original sequence is mapped to the sequence. In the window subsequence, the window will continue to move backward to introduce new sites; as the window continues to move backward, there will eventually be a time point when the window can no longer introduce new sites, and then proceed to step (d).

[0025] (d) Mapping window size The value is reduced by 1, and the original sequence is perturbed again in the manner of step (b);

[0026] (e) when When the value decreases to 1, the perturbation process ends; the original symbolic sequence is transformed into a perturbed symbolic sequence.

[0027] Furthermore, throughout the process, the sequence site mapping probabilities are as follows:

[0028]

[0029] In the formula: This indicates moving backwards with the current original sequence position as a reference. The probability of mapping a site to a mapping sequence; The privacy budget represents the localized differential privacy; Indicates the window size; Indicates what can be obtained The total set of values; This means removing the smallest one. Value after A subset of; noting that with the current original sequence position as the reference frame, there is a possibility of forward mapping. To avoid confusion in the mapping direction, when forward mapping, take The value is negative.

[0030] Furthermore, in step four, when the required charge / discharge amount exceeds the parameter limit, the maximum actual energy that can be provided is taken as the final energy storage charge / discharge amount, specifically mathematically represented as follows:

[0031]

[0032] In the formula: This refers to the corrected energy storage charge and discharge rates; This refers to the initial energy storage charging and discharging point volume; This represents the maximum charging rate for energy storage. This represents the maximum discharge rate of the stored energy. Let be the remaining amount of energy stored at time t.

[0033] The beneficial effects of this invention are as follows:

[0034] (1) The present invention takes into account the strong time sequence of the user's energy consumption curve itself. First, the load characteristics are extracted by using the time series symbol aggregation method, and then time series random mapping is performed on this basis. This hides the real load data and disturbs the potential user electricity consumption behavior in the load curve. The resulting energy storage charging and discharging behavior can retain the statistical characteristics of the original energy consumption curve while protecting the user's load time sequence privacy.

[0035] (2) The present invention can protect users’ electricity privacy from the perspective of load timing without destroying the original load statistical characteristics.

[0036] (3) The present invention uses a high-confidence method to determine the energy storage parameters required by users. These parameters can reduce the impact of physical constraints of energy storage devices on privacy protection to a certain extent, and provide a reference for the configuration of user-side energy storage. Attached Figure Description

[0037] Figure 1 A flowchart illustrating a localized differential privacy protection method for energy consumption curves that takes into account load timing characteristics.

[0038] Figure 2 This is a schematic diagram of load time series feature extraction based on the time series symbol aggregation method. Detailed Implementation

[0039] The present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments. The purpose and effects of the present invention will become clearer. It should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0040] like Figure 1 As shown, the localized differential privacy protection method for energy consumption curves that considers load timing characteristics according to an embodiment of the present invention includes the following steps:

[0041] Step 1: Extract features from user energy consumption curves using time-series symbol aggregation. Time-series symbol aggregation effectively extracts temporal variation features from energy consumption curves, which is crucial for protecting the privacy of load temporal features. The method pre-determines the number of characters and states, and based on this, transforms the user energy consumption curve from a time series into a discrete character sequence, where each character represents a specific load temporal feature.

[0042] In scenarios where user energy consumption curve privacy is protected, the specific process is as follows:

[0043] (a) Normalize the energy consumption curve to transform the numerical range in the time series into [0,1], thereby reducing the impact of adverse data on the series. The normalization method is as follows:

[0044]

[0045] In the formula: The value is the normalized value; The first time series One value; The maximum value in the sequence; The minimum value of the sequence;

[0046] (b) The energy consumption curve is approximated by load segmentation aggregation, and the amplitude of the decrease in the same time interval is replaced by the average value. The calculation method is as follows:

[0047]

[0048] In the formula: For the first The average value of the data segment; For the first The time points of each load data point; An index for normalized load data; This is an index of the load data after approximate transformation through load segmentation and aggregation;

[0049] (c) The energy consumption curve is symbolized as a discrete string. That is, the energy consumption curve after load segmentation and aggregation approximation is divided into intervals according to the amplitude. Each character represents a certain amplitude range. The relationship between the converted amplitude and the character is as follows:

[0050]

[0051] In the formula: The i-th character in the character list; The character in the preset p-th segment has a numerical range of 1. .

[0052] After three steps of data processing, an energy consumption curve can be represented by a string of symbols. To express.

[0053] Step 2: Utilize a variable window random mapping mechanism to achieve temporal perturbation of the symbolic sequence; determine the sequence site mapping probability with a time window span of k, and perturb the formed symbolic sequence temporally through randomized sequence site mapping, transforming it into a perturbed symbolic sequence. The symbolic sequence is then numerically converted into a new user energy consumption curve. The new energy consumption curve enables localized differential privacy protection compared to the original energy consumption curve.

[0054] Step two specifically includes the following sub-steps:

[0055] (a) Assume that the original energy consumption curve forms an original symbol sequence after being processed by the time series symbol aggregation method. At the same time, there exists a mapping sequence with the same length as the symbol sequence. In the initial stage, all points in the mapping sequence are empty, and a time span is... The window is placed at the beginning of both the original sequence and the mapped sequence;

[0056] (b) Initial random mapping, the original sequence subsequence located within the window First site Mapped to a certain probability Window subsequence of a sequence At any site; subsequently, the mapped sites in the window subsequence of the original sequence are discarded, the window is shifted forward to add new sites and form a new window subsequence, and the number of sequence sites in the window remains unchanged. Mapped sites in the window subsequence of the mapped sequence are discarded, and the window moves one position to the right, while the number of sites in the window remains the same. ;

[0057] (c) In each subsequent random mapping, referring to step (b), a site in the window subsequence of the original sequence is mapped to the sequence. In the window subsequence, the window will continue to move backward to introduce new sites; as the window continues to move backward, there will eventually be a time point when the window can no longer introduce new sites, and then proceed to step (d).

[0058] (d) Mapping window size The value is reduced by 1, and the original sequence is perturbed again in the manner of step (b);

[0059] (e) when When the value decreases to 1, the perturbation process ends; the original symbolic sequence is transformed into a perturbed symbolic sequence.

[0060] Throughout the process, the sequence site mapping probabilities are as follows:

[0061]

[0062] In the formula: This indicates moving backwards with the current original sequence position as a reference. The probability of mapping a site to a mapping sequence; The privacy budget represents the localized differential privacy; Indicates the window size; Indicates what can be obtained The total set of values; This means removing the smallest one. Value after A subset of; noting that with the current original sequence position as the reference frame, there is a possibility of forward mapping. To avoid confusion in the mapping direction, when forward mapping, take The value is negative.

[0063] Step 3: To ensure the operational efficiency of user-side energy storage devices, a high-confidence energy storage parameter selection method is used to determine the required energy storage capacity for different residential users. Steps 1 and 2 are repeated to process one month's historical data from each user, obtaining a set of minimum capacity and rate parameters that can meet all mapping requirements. This process is repeated multiple times, and then the parameters that meet the confidence threshold are selected as the required energy storage device parameters for each user. For example, this is repeated 100 times, and the 80th value in the sorted data set is taken as the required energy storage device parameter for each user.

[0064] Step Four: Utilize the charging and discharging behavior of the energy storage device required by the user, as determined in Step Three, to transform the original energy consumption curve into the disturbed energy consumption curve. Even with energy storage parameters determined with high confidence, energy limitations are inevitable. To address this issue, a simplified processing method is adopted. When the required charging and discharging amount exceeds the parameter limits, the maximum actual energy provided is used as the final energy storage charging and discharging amount. The specific mathematical representation is as follows:

[0065]

[0066] In the formula: This refers to the corrected energy storage charge and discharge rates; This refers to the initial energy storage charging and discharging point volume; This represents the maximum charging rate for energy storage. This represents the maximum discharge rate of the stored energy. Let be the remaining amount of energy stored at time t.

[0067] The following is a specific example to further illustrate the construction of the variable window random map mentioned in this invention.

[0068] (1) First, assume that the energy consumption curve data sequence of a certain user on a certain day is as follows:

[0069] [0.2334,0.2156,0.1608,0.2282,0.3794,1.2641,0.4104,0.4701,0.4824,0.4322,0.5873,0.5617,0.6168,0.6386,0.5565,0.5626,0.573,0.3989,0.4562,0.4131,0.3937,0.3887,1.2619,1.35 23, 1.299, 0.5765, 0.3718, 0.3854, 0.5026, 1.3363, 0.3617, 0.3654, 0.3605, 0.4286, 0.3631, 0.3565, 0.3585, 0.3585, 0.3957, 0.4153, 0.3524, 0.3151, 1.2875, 1.1654, 1.0361, 0.2911, 0.2937, 0. 3585, 0.3167, 0.4575, 0.5674, 0.3724, 0.1823, 0.1814, 0.1906, 0.2208, 0.22, 0.2259, 0.2368, 0.2294, 1.1036, 1.09, 2.3525, 1.8093, 0.1681, 0.1683, 0.3642, 0.226, 0.2239, 0.2294, 0.1456, 0.28 11,0.2114,0.2122,0.2117,0.2121,0.2915,0.2009,0.1526,0.1591,0.1507,0.2189,0.1523,0.1517,0.1587,0.1499,0.2243,0.1526,0.2131,0.1447,0.1517,0.2174,0.1537,0.1558,0.2107).

[0070] The symbol list length for the time series symbol aggregation method is set to 7, and each load segment contains 3 time points. After extracting the load time series features using this method, the resulting symbolized sequence is:

[0071] [a,c,c,c,e,c,c,c,a,c,a,a,a,a,a,a,c,a,a,a,a,a,a,e,a,a,a,a,a,a,a,c,a].

[0072] A schematic diagram of load time-series feature extraction based on time-series symbol aggregation method is shown below. Figure 2 As shown.

[0073] (2) The symbolized sequence is perturbed using a variable window random mapping mechanism. Assume the obtained user original energy consumption curve symbolized sequence... for Window size The value is 3, representing the initial phase mapping sequence. Empty. The sequence at the start of the first mapping. window subsequence The first site Can be mapped to Window subsequence of a sequence At any given site. Based on the designed site mapping probability, the site... numerical value Mapped to The probability at that location is And mapped to and The probability of all is Set the value with the higher probability in the mapping probability as... The value with a lower probability is set to Assuming that after the first mapping, the numerical value... Mapped to At this time, the mapping sequence Become The window moves backward. Abandoned by the window Entering the window, the window subsequence of the mapping sequence starts from... Become The window subsequence of the original sequence is composed of Become The next site to be mapped is the second site of the original sequence. The value is At this point, it is mapped to... The probability is still 1 The probability of mapping to the other two sites is 1. After mapping is completed again with the same probability, new sequence sites are added to the window, and mapped sequence sites are removed from the window, and so on until the mapping is finished.

[0074] Assuming the window size k is 10 and the privacy budget ε is 1+ln2, the symbol sequence in (1) is randomly mapped using a variable window random mapping mechanism to obtain a new symbol sequence. Based on the original data corresponding to the symbol sequence, a new payload sequence is generated.

[0075] [0.2484,0.2482,0.3996,0.3977,0.4768,0.3879,0.3905,0.3513,0.4037,0.6201,0.6176,0.462,1.316,1.1222,0.4195,0.4316,0.4932,0.422,0.4697,0.3869,0.3867,0.3422,0.292,0.186,0 .2692,0.2803,0.1423,0.2219,0.1926,1.2877,0.2583,0.2585,0.2042,0.2282,0.2077,0.1644,0.2201,0.2162,0.2161,0.2603,0.3347,0.3685,0.2285,0.2262,0.2115,0.203,0.2047,0.2045, 0.1318,0.2032,0.2055,0.2966,0.2583,0.1935,0.2362,0.2302,1.8132,0.2716,0.2209,0.1882,0.1783,0.1808,0.2522,0.2846,0.2872,0.2398,0.292,0.3102,0.2767,0.3054,0.3134,0.2429 ,0.1541,0.1546,0.1548,0.22,0.3057,0.2412,0.1552,0.1549,0.2194,0.2401,0.3232,0.2621,0.4975,0.5054,0.3755,0.1737,0.174,0.2385,0.2393,0.322,0.2651,0.1553,0.1569,1.0944).

[0076] (3) The two symbol sequences above are rounded to two decimal places. The conversion from the original load sequence to the new load sequence is achieved through the charging and discharging behavior of the household energy storage device. When the required charging and discharging amount exceeds the parameter limit, the maximum energy that can actually be provided is used as the final charging and discharging amount of the stored energy. The final load sequence is as follows:

[0077] [0.22,0.19,1.29,0.39,0.35,0.4,0.62,0.62,0.46,1.32,1.12,0.42,0.47,0.39,0.39,0.25,0.25,0.4,0.13,0,0.21,0.23,0.21,0.16,0.23,0.23,0.21,0.34,0.29,0.19,0.4,0.48,0.39,0.3,0.26,0.19,0.43,0.49,0.42,0.22,0.22,0.22,0.18,0.18,0.25,0.27,0.22,0.19, 0.29,0.31,0.2,0.2,0.2,0.2,0.26,0.26,0.2,0.22,0.31,0.24,0.17,0.17,0.24,0.28,0.29,0.24,0.27,0.28,0.14,0.24,0.23,1.81,0.16,0.15,0.22,0.24,0.32,0.27,0.15,0.15,0.15,0.3,0.47,0.42,0.21,0.2,0.32,0.24,0.32,0.26,0.26,0.33,0.37,0.26,0.35,0.21).

[0078] It will be understood by those skilled in the art that the above descriptions are merely preferred examples of the invention and are not intended to limit the invention. Although the invention has been described in detail with reference to the foregoing examples, those skilled in the art can still modify the technical solutions described in the foregoing examples or make equivalent substitutions for some of the technical features. All modifications and equivalent substitutions made within the spirit and principles of the invention should be included within the scope of protection of the invention.

Claims

1. A localized differential privacy protection method for energy consumption curves considering load timing characteristics, characterized in that, Includes the following steps: Step 1: Symbolize the original user energy consumption curve using the time series symbol aggregation method, transforming the user energy consumption curve from a time series into a discrete character sequence. Each character in the sequence represents a certain load time series feature, thereby effectively extracting the variation features in the energy consumption curve and providing a basis for achieving privacy protection of load time series features. Step 2: Using the variable window random mapping mechanism, the symbolic original user energy consumption curve is transformed into a perturbed symbol sequence. The mapping probability of the sequence site with a time window span of k is determined. The symbolic sequence is temporally perturbed by the randomized mapping method of the sequence site, and transformed into a perturbed symbol sequence. The symbol sequence is then numerically converted into a new user energy consumption curve to ensure that the sequence can achieve localized differential privacy protection of the original sequence. Step 3: Repeat steps 1 and 2 to process the user's historical data for one month to obtain a set of minimum energy storage capacity parameters and rate parameters that can meet all mapping requirements. Then, select the parameters that meet the confidence threshold from these parameters as the energy storage device parameters required by the user. Step 4: Utilize the charging and discharging behavior of the energy storage device required by the user as determined in Step 3 to transform the original energy consumption curve into the disturbed energy consumption curve; when the required charging and discharging amount exceeds the parameter limit, the maximum actual energy provided is used as the final charging and discharging amount provided by the energy storage. Step two specifically includes the following sub-steps: (a) Assume that the original energy consumption curve forms an original symbol sequence after being processed by the time series symbol aggregation method. At the same time, there exists a mapping sequence with the same length as the symbol sequence. In the initial stage, all points in the mapping sequence are empty, and a time span is... The window is placed at the beginning of both the original sequence and the mapped sequence; (b) Initial random mapping, the original sequence subsequence located within the window First site Mapped to a certain probability Window subsequence of a sequence At any site; subsequently, the mapped sites in the window subsequence of the original sequence are discarded, the window is shifted forward to add new sites and form a new window subsequence, and the number of sequence sites in the window remains unchanged. ; Mapped sites in the window subsequence of the mapped sequence are discarded, and the window moves one position to the right synchronously, while the number of sites in the window remains the same. ; (c) In each subsequent random mapping, referring to step (b), a site in the window subsequence of the original sequence is mapped to the sequence. Within the window subsequence, and the window will continue to move forward to introduce new sites; As the window continues to move backward, there eventually comes a point in time when the window can no longer introduce new sites, at which point the process proceeds to step (d). (d) Mapping window size The value is reduced by 1, and the original sequence is perturbed again in the manner of step (b); (e) when When the value decreases to 1, the perturbation process ends; the original symbolic sequence is transformed into a perturbed symbolic sequence. Throughout the process, the sequence site mapping probabilities are as follows: ； In the formula: This indicates moving backwards with the current original sequence position as a reference. The probability of mapping a site to a mapping sequence; The privacy budget represents the localized differential privacy; Indicates the window size; Indicates what can be obtained The total set of values; This means removing the smallest one. Value after A subset of; noting that with the current original sequence position as the reference frame, there is a possibility of forward mapping. To avoid confusion in the mapping direction, when forward mapping, take The value is negative.

2. The localized differential privacy protection method for energy consumption curves considering load timing characteristics according to claim 1, characterized in that, Step one includes the following sub-steps: (a) Normalize the energy consumption curve to transform the numerical range in the time series into [0,1], thereby reducing the impact of bad data on the series; the normalization method is as follows: ； In the formula: The value is the normalized value; The first time series One value; The maximum value in the sequence; The minimum value of the sequence; (b) The energy consumption curve is approximated by load segmentation aggregation, and the average value is used to replace the amplitude of the decrease in the same time interval. The calculation method is as follows: ； In the formula: For the first The average value of the data segment; For the first The time points of each load data point; An index for normalized load data; This is an index of the load data after approximate transformation through load segmentation and aggregation; (c) The energy consumption curve is symbolized as a discrete string. That is, the energy consumption curve after load segmentation and aggregation approximation is divided into intervals according to the amplitude. Each character represents a certain amplitude range. The relationship between the converted amplitude and the character is as follows: ； In the formula: The i-th character in the character list; The character in the preset p-th segment has a numerical range of 1. .

3. The localized differential privacy protection method for energy consumption curves considering load timing characteristics according to claim 1, characterized in that, In step four, when the required charge / discharge amount exceeds the parameter limit, the maximum actual energy that can be provided is taken as the final energy storage charge / discharge amount, specifically represented mathematically as follows: ； In the formula: This refers to the corrected energy storage charge and discharge rates; This refers to the initial energy storage charging and discharging point volume; This represents the maximum charging rate for energy storage. This represents the maximum discharge rate of the stored energy. Let be the remaining amount of energy stored at time t.

Images