PVPS component and system reliability engineering works to define the PVPS probability of making the indicated value such as energy or revenue, also at a given statistical confidence level for an estimate. This needs to be assessed properly as an accurate levelized cost of energy (LCOE) results from identifying and acting on a set of quantifiable metrics based upon real measured data of actual plants under the widest variety of real site conditions. In many instances, the use of P numbers (which stands for “percentile”) may not be clearly understood and as a result, inappropriate conclusions drawn which have a financial result. P values are used to establish the confidence that one can require to provide the assurance that the item will meet specification. A P50 value, for example, provides that there is a 50 % confidence in the value used in reliability predictions. This value of confidence translates to the median of the population or in other words, it is equivalent to a coin toss on whether the value is valid. It is better to have a higher confidence that the system will work to specification. For reliability metrics, this is typically defined as being a P90 or P95 values. This level of confidence significantly characterizes financial and technical risk plant availability.
The failure rates and mode become important for predicting future failures. In a worst case, significant wear out failures may be indicative of serial failures and attention is warranted. A needed caution is the components may have multiple failure modes and root cause analyses may be useful discerning the failure modes.
The LCOE calculations may not adequately include all the relevant costs, i.e. all-in costs, and risks which create further uncertainty. That uncertainty has a high probability of coming to inaccurate conclusions and choices.
Ideally, the owners, maintainers and operators should look for reliability issues early in the concept, system, and hardware and software design engineering efforts. Otherwise, the defects in software code and poor design or weak components will manifest themselves in a multitude of unexpected failures resulting in unwanted and unexpected risks and costs.
In addition, there is another issue that is a by-product of unexpected costs. Organizational angst is the result of not addressing issues at specification prior to design that in turn results in organizational effort, time, and expense in the solving of problems (often originally simple) that become quite complicated after the plant has been built. Because this effort may not be adequately budgeted, and places additional stress on the organization, it tends to have a negative impact on the human performance of scope and adds risk to the PVPS performance.
Without analysis of accurate field data and metrics, there are a series of negative results that include unidentified or unexpected levels of plant failures and degradation. Lack of ongoing (from concept to end-of-life project phases) reliability analyses, the results of inaction raise unaddressed costs, risks, reduced plant capacity and capability, and potential for plant derating. All these issues could potentially result in substantial negative financial impacts to the owners, insurers, users and/or operators.
Reliability of a PVPS requires a comprehensive approach to identify, maintain, correct, and understand costs. Some critically necessary specific gaps for the PV industry need advancement:
a) A standard way to define failure statistics for PV, for PV components and specifically PV modules where failure can be either catastrophicor degradation-driven. This can be accomplished by a bottoms-up fault tree nodal model with further guidance on how each of the nodal distributions can be derived qualitatively.
b) Defining a common nomenclature of describing failures in the field so that failure statistics can be gathered and analysed (i.e., failure coded or word search capability). Further there needs to be coordination between the various stakeholders to standardize data capture in a format that allows for meta-analysis. Different levels of data can be used for different or enhanced understanding of reliability issues depending on available technology and installed capability. Improvement in monitoring is assumed but there is a need to create standardization criteria, and details on data capture.
c) Defining a standard for how operational failure data is classified, root cause identified, and reported to aid objective b) with guidance or criteria established or cited.
Reliable systems, processes, and procedures produce energy more safely at a consistently lower cost while reducing waste, unnecessary labour, unplanned O&M, and unnecessary organizational angst while providing additional actionable information to continually build and operate better, higher producing and safer plants.
An obvious concern is that the system appears imposing at first sight. It is not the intention that the effort be a greater cost than its benefits. The resultant specifications and design shall fit the business /financial needs of the project. The cost of ensuring reliability needs to be weighed against the costs of not ensuring reliability at achievable levels. The types of data and commitment to data collection, however, should be tempered while addressing the initial and future data requirements. The Pareto techniques allow insights to be gained on the vital few as per the 80/20 % rule (see 7.11). However, much data needs to be collected and this provides references to other documents that address data.
